阅读说明：本技术 用于检测周期性信号伪影的概率熵 (Probability entropy for detecting periodic signal artifacts ) 是由 E·J·潘肯 J·C·杰克逊 Y·肖 C·L·普利亚姆 于 2020-04-17 设计创作，主要内容包括：本发明公开了用于使用概率熵来选择具有较少伪影的电极以用于控制适应性电神经刺激的技术。在一个示例中,多个电极感测患者的脑的生物电信号。处理电路针对在多个电极中的相应电极处感测到的每个生物电信号确定生物电信号的概率熵值。该处理电路将生物电信号的相应概率熵值中的每个概率熵值与相应熵阈值进行比较,以及基于比较来选择多个电极中的电极的子集。该处理电路基于经由电极的子集中的相应电极感测到的生物电信号并且排除多个生物电信号中的经由不在电极的子集中的相应电极感测到的生物电信号来控制向患者递送电刺激治疗。(Techniques for selecting electrodes with fewer artifacts for controlling adaptive electrical nerve stimulation using probabilistic entropy are disclosed. In one example, a plurality of electrodes sense bioelectrical signals of a brain of a patient. The processing circuitry determines a probability entropy value for the bioelectrical signals for each bioelectrical signal sensed at a respective electrode of the plurality of electrodes. The processing circuit compares each of the respective probability entropy values of the bioelectrical signal to a respective entropy threshold, and selects a subset of the electrodes of the plurality of electrodes based on the comparison. The processing circuitry controls delivery of electrical stimulation therapy to the patient based on the bioelectrical signals sensed via respective electrodes of the subset of electrodes and excludes bioelectrical signals of the plurality of bioelectrical signals that are sensed via respective electrodes that are not in the subset of electrodes.)

1. An implantable medical device, the implantable medical device comprising:

sensing circuitry configured to sense a plurality of bioelectrical signals of a brain of a patient via a plurality of electrodes; and

a processing circuit configured to:

determining, for each bioelectrical signal of the plurality of bioelectrical signals sensed at a respective electrode of the plurality of electrodes, a probability entropy value for the bioelectrical signal;

comparing each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold;

selecting a subset of electrodes of the plurality of electrodes based on the comparison; and

controlling delivery of electrical stimulation therapy to the patient based on the bioelectrical signals of the plurality of bioelectrical signals sensed via respective electrodes of the subset of electrodes and excluding the bioelectrical signals of the plurality of bioelectrical signals sensed via respective electrodes not in the subset of electrodes.

2. The implantable medical device of claim 1, wherein to determine the probability entropy value of the bioelectrical signal, the processing circuitry is further configured to determine a probability distribution of entropy of the bioelectrical signal over a period of time,

wherein a signal component of the bioelectric signal having periodic behavior is indicative of reduced entropy and wherein a signal component of the bioelectric signal not having periodic behavior is indicative of increased entropy.

3. The implantable medical device of any one of claims 1-2, wherein to determine the probability entropy value for each bioelectrical signal, the processing circuitry is further configured to:

determining that one or more signal components of a first bioelectric signal of the plurality of bioelectric signals exhibit periodic behavior;

determining a first probability entropy value for the first bioelectrical signal in response to determining that the one or more signal components of the first bioelectrical signal exhibit periodic behavior;

determining that one or more signal components of a second bioelectric signal of the plurality of bioelectric signals exhibits aperiodic behavior; and

determining a second probability entropy value for the second bioelectric signal in response to determining that the one or more signal components of the second bioelectric signal exhibit aperiodic behavior;

wherein the first probability entropy value indicates entropy in the first bioelectric signal and the second probability entropy value indicates entropy in the second bioelectric signal, and

wherein the first probability entropy value and the second probability entropy value indicate that the first bioelectric signal exhibits less entropy than the second bioelectric signal.

4. The implantable medical device of any one of claims 1-3, wherein to determine the probability entropy value of the bioelectrical signal, the processing circuitry is further configured to determine a statistical measure of randomness over a period of time.

5. The implantable medical device of claim 4,

wherein to determine the probability entropy value of the bioelectric signal, the processing circuitry is further configured to determine a statistical measure of randomness of spectral power over a plurality of frequency bands of the bioelectric signal,

wherein to compare each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold, the processing circuitry is further configured to compare each statistical measure of randomness of spectral power over the plurality of frequency bands of the bioelectric signal to an entropy threshold, and

wherein to select a subset of electrodes of the plurality of electrodes based on the comparison, the processing circuitry is further configured to select each electrode of the plurality of electrodes that satisfies the following condition: for each of the electrodes, the statistical measure of randomness of spectral power over the plurality of frequency bands of the respective bioelectrical signal sensed via the electrode is greater than the respective entropy threshold.

6. The implantable medical device of any one of claims 1-5,

wherein to compare each of the respective probability entropy values of the bioelectrical signal to a respective entropy threshold, the processing circuitry is further configured to:

determining a rate at which the amplitude of each bioelectric signal exceeds a threshold limit; and

comparing the rate at which the amplitude of the bioelectrical signal exceeds the threshold limit to a respective rate threshold; and is

Wherein to select a subset of electrodes of the plurality of electrodes based on the comparison, the processing circuitry is further configured to select each electrode of the plurality of electrodes that satisfies the following condition: for each of the electrodes, a rate at which an amplitude of the respective bioelectrical signal sensed via the electrode exceeds the threshold limit is less than the respective rate threshold.

7. The implantable medical device of claim 6, wherein each threshold limit is a quartile distance of the sensed amplitude of the respective bioelectric signal.

8. The implantable medical device of any one of claims 1-7,

wherein to compare each of the respective probability entropy values of the bioelectrical signal to a respective entropy threshold, the processing circuitry is further configured to:

determining an entropy of a time interval between instances where the amplitude of the bioelectrical signal exceeds a threshold limit; and

comparing the entropy of the time interval with the respective entropy threshold, and

wherein to select a subset of electrodes of the plurality of electrodes based on the comparison, the processing circuitry is further configured to select each electrode of the plurality of electrodes that satisfies the following condition: for said each electrode, said entropy of said time interval is greater than said respective entropy threshold.

9. The implantable medical device of claim 8, wherein each threshold limit is a first quartile of the sensed amplitude of the respective bioelectric signal.

10. The implantable medical device of any one of claims 1-9, wherein to compare each of the respective probability entropy values of the bioelectrical signal to a respective entropy threshold and select the subset of the electrodes of the plurality of electrodes based on the comparison, the processing circuitry is configured to execute a machine learning system configured to:

processing each of the respective probability entropy values of the bioelectrical signal to identify one or more electrodes of the plurality of electrodes, an

Selecting the identified one or more electrodes as a subset of the electrodes in the plurality of electrodes.

11. The implantable medical device of any one of claims 1-10, wherein to compare each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold, the processing circuitry is configured to:

determining a normalization of each bioelectrical signal; and

comparing the normalized probability entropy of the bioelectrical signal with the respective entropy threshold.

12. The implantable medical device of any one of claims 1-11, wherein the processing circuitry is further configured to output, for each electrode not in the subset of electrodes, an indication that an artifact is present in the respective bioelectric signal sensed via the electrode.

13. A method performed by the implantable medical device of any one of claims 1-12.

14. A system comprising the implantable medical device of any one of claims 1-12.

15. An implantable medical device, the implantable medical device comprising:

means for sensing a plurality of bioelectrical signals of a brain of a patient via a plurality of electrodes; and

means for determining, for each bioelectric signal of the plurality of bioelectric signals sensed at a respective electrode of the plurality of electrodes, a probability entropy value for the bioelectric signal;

means for comparing each of the respective probability entropy values of the bioelectrical signal to a respective entropy threshold;

means for selecting a subset of electrodes of the plurality of electrodes based on the comparison; and

means for controlling delivery of electrical stimulation therapy to the patient based on the bioelectrical signals of the plurality of bioelectrical signals sensed via respective electrodes of the subset of electrodes and excluding the bioelectrical signals of the plurality of bioelectrical signals sensed via respective electrodes not in the subset of electrodes.

Technical Field

The present disclosure relates generally to electrical stimulation therapy.

Background

The medical device may be external or implanted, and may be used to deliver electrical stimulation therapy to various tissue sites of a patient to treat a variety of symptoms or conditions, such as chronic pain, tremor, parkinson's disease, other movement disorders, epilepsy, urinary or fecal incontinence, sexual dysfunction, obesity, or gastroparesis. The medical device may deliver electrical stimulation therapy via one or more leads that include electrodes located near a target location associated with the brain, spinal cord, pelvic nerves, peripheral nerves, or gastrointestinal tract of the patient. Thus, electrical stimulation may be used for different therapeutic applications, such as adaptive deep brain stimulation (aDBS), Spinal Cord Stimulation (SCS), pelvic stimulation, gastric stimulation, Peripheral Nerve Field Stimulation (PNFS), electroencephalography (EEG), corticography (ECoG), Electromyography (EMG), or for performing biopotential recordings of other channels of a patient.

The clinician may select values for a plurality of programmable parameters in order to define the electrical stimulation therapy to be delivered to the patient by the implantable stimulator. For example, the clinician may select one or more electrodes for delivering stimulation, the polarity of each selected electrode, the voltage or current amplitude, the pulse width, and the pulse frequency as stimulation parameters. A set of parameters, such as a set of parameters including electrode combination, electrode polarity, amplitude, pulse width, and pulse frequency, may be referred to as a program in the sense of defining an electrical stimulation therapy to be delivered to a patient.

Disclosure of Invention

Bioelectric signals sensed from a patient, such as Local Field Potentials (LFPs), EEG, ECoG, or EMG, may be used as biomarkers or input signals for a control system for therapy delivery, such as aDBS. For example, the LFP signal may be used as a biomarker for controlling one or more parameters of the electrical stimulation therapy delivered to the patient. However, if the recorded bioelectric signals are contaminated by artifacts, the clinical effectiveness of the recorded bioelectric signals may be compromised. A variety of factors, such as Electrocardiogram (ECG) signals or repetitive motion, may distort or introduce artifacts in the recorded bioelectric signals. The amplitude of these artifacts is typically variable across recordings, and therefore medical devices may be limited in detecting the artifacts by conventional algorithms. However, the artifact may be periodic in nature (e.g., heartbeat, pacing therapy), and the periodicity may be used to construct the detection algorithm. As used herein, periodicity refers to a pattern or order in a signal that has a lower entropy (e.g., randomness) than a bioelectric signal (such as neuronal LFP activity) that may exhibit more random (e.g., stochastic) properties. Thus, the entropy of one or more features of the recorded bioelectrical signal may be used as a salient feature to identify periodic artifacts.

Techniques are disclosed for distinguishing electrodes capable of sensing clean bioelectric signals from electrodes contaminated by artifacts using probabilistic entropy. In some examples, the techniques may be used to verify that the bioelectric signals sensed by the recording electrodes are of sufficient quality to be used as biomarkers for controlling the aDBS therapy. In some examples, the probability entropy may be used as an indicator of periodic artifacts (such as ECG) present in the recorded LFP signals of the patient's brain. In one example, a plurality of electrodes sense bioelectrical signals of a brain of a patient. The processing circuitry determines a probability entropy value for the bioelectrical signals for each bioelectrical signal sensed at a respective electrode of the plurality of electrodes. The processing circuit compares each of the respective probability entropy values of the bioelectrical signal to a respective entropy threshold and selects a subset of the electrodes of the plurality of electrodes based on the comparison. Thus, the processing circuitry may use the probability entropy values of the bioelectrical signals to improve selection of electrodes for sensing the bioelectrical signals of the patient or for delivering therapy to the patient. For example, the processing circuitry controls delivery of electrical stimulation therapy to the patient based on the bioelectrical signals sensed via respective electrodes of the subset of electrodes and excludes bioelectrical signals of the plurality of bioelectrical signals sensed via respective electrodes not in the subset of electrodes. As another example, the processing circuitry senses one or more bioelectrical signals of the patient based on bioelectrical signals sensed via respective electrodes of the subset of electrodes and excludes bioelectrical signals of the plurality of bioelectrical signals sensed via respective electrodes not in the subset of electrodes.

Accordingly, the techniques disclosed herein may provide enhanced accuracy in the identification of artifacts in the electrodes. For example, the techniques of this disclosure may detect artifacts that may otherwise be difficult to detect using conventional artifact detection methods, such as artifacts with variable signal amplitudes across multiple recordings. Thus, by identifying and eliminating measurements from artifact-contaminated electrodes, the techniques of this disclosure may provide greater reliability in an acdb system. For example, the techniques of this disclosure may increase the following possibilities: the signals sensed by the electrodes and used as biomarkers of the aDBS accurately reflect the true bioelectric signals and avoid erroneous measurements that may adversely affect the therapy provided to the patient. Accordingly, the techniques disclosed herein may provide patients with more effective aDBS therapy than conventional systems.

In one example, the present disclosure describes a method comprising: sensing a plurality of bioelectrical signals of a brain of a patient via a plurality of electrodes; determining, by the processing circuitry and for each bioelectrical signal of the plurality of bioelectrical signals sensed at a respective electrode of the plurality of electrodes, a probability entropy value for the bioelectrical signal; comparing, by the processing circuit, each of the respective probability entropy values of the bioelectric signals to a respective entropy threshold; and selecting, by the processing circuitry and based on the comparison, a subset of the electrodes of the plurality of electrodes; and controlling, by the processing circuitry and based on and excluding ones of the plurality of bioelectrical signals sensed via respective electrodes of the subset of electrodes, delivery of the electrical stimulation therapy to the patient.

In another example, the present disclosure describes an implantable medical device comprising: a plurality of electrodes; sensing circuitry configured to sense a plurality of bioelectrical signals of a brain of a patient via a plurality of electrodes; and processing circuitry configured to: determining a probability entropy value for the bioelectric signals for each bioelectric signal of the plurality of bioelectric signals sensed at a respective electrode of the plurality of electrodes; comparing each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold; selecting a subset of electrodes of the plurality of electrodes based on the comparison; and controlling delivery of electrical stimulation therapy to the patient based on and excluding ones of the plurality of bioelectrical signals sensed via respective electrodes of the subset of electrodes.

In another example, the present disclosure describes a system comprising: an implantable medical device, comprising: a plurality of electrodes; sensing circuitry configured to sense a plurality of bioelectrical signals of a brain of a patient via a plurality of electrodes; and processing circuitry configured to: determining, for each bioelectric signal of a plurality of bioelectric signals sensed at a respective electrode of a plurality of electrodes, a probability entropy value for the bioelectric signal; comparing each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold; selecting a subset of electrodes of the plurality of electrodes based on the comparison; and controlling delivery of therapy, such as electrical stimulation therapy, to the patient based on and excluding ones of the plurality of bioelectrical signals sensed via respective electrodes of the subset of electrodes.

The details of one or more examples of the techniques of this disclosure are set forth in the accompanying drawings and the description below.

Drawings

Fig. 1 is a conceptual diagram illustrating an exemplary system including an Implantable Medical Device (IMD) configured to deliver an adaptive DBS to a patient, according to an example of the techniques of this disclosure.

Fig. 2 is a block diagram of the example IMD of fig. 1 for delivering adaptive DBS therapy in accordance with an example of the techniques of this disclosure.

Fig. 3 is a block diagram of the external programmer of fig. 1 for controlling delivery of adaptive DBS therapy in accordance with an example of the techniques of this disclosure.

Fig. 4 is a graphical representation of a sensed bioelectrical signal of a patient.

Fig. 5 is a flow chart illustrating exemplary operations in accordance with the techniques of this disclosure.

Fig. 6 is a flow chart illustrating exemplary operations in accordance with the techniques of this disclosure.

Fig. 7 is a flow chart illustrating exemplary operations in accordance with the techniques of this disclosure.

Fig. 8 is a flow chart illustrating exemplary operations in accordance with the techniques of this disclosure.

Like reference numerals refer to like elements throughout the drawings and the description.

Detailed Description

Fig. 1 is a conceptual diagram illustrating an exemplary system 100 including an Implantable Medical Device (IMD)106 configured to deliver adaptive deep brain stimulation to a patient 112. DBS may be adaptive in the sense that IMD106 may adjust, increase, or decrease the magnitude of one or more parameters of DBS in response to changes in patient motion or movement, the severity of one or more symptoms of the patient's disease, the presence of one or more side effects due to DBS, or one or more sensed bioelectric signals of the patient, etc. For example, one or more sensed signals of the patient may be used as control signals such that IMD106 correlates the magnitude of one or more parameters of the electrical stimulation to the magnitude of one or more characteristics of one or more sensed bioelectrical signals. IMD106 may deliver electrical stimulation therapy having one or more parameters, such as voltage or current amplitudes, adjusted in response to the magnitude of one or more characteristics of one or more sensed bioelectrical signals.

System 100 may be configured to treat a patient disorder, such as a movement disorder, a neurodegenerative injury, a mood disorder, or epilepsy in patient 112. Patient 112 is typically a human patient. However, in some cases, treatment system 100 may be applied to other mammalian or non-mammalian, non-human patients. Although primarily referenced herein to movement disorders and neurodegenerative impairments, in other examples, treatment system 100 can provide treatment to manage symptoms of other patient disorders, such as, but not limited to, epilepsy (e.g., epilepsy) or mood (or psychological) disorders (e.g., Major Depressive Disorder (MDD), bipolar disorder, anxiety disorder, post-traumatic stress disorder, pain, spasticity, incontinence, dysthymic disorder, and Obsessive Compulsive Disorder (OCD)). At least some of these disorders may be manifested as one or more patient motor behaviors. As described herein, a movement disorder or other neurological impairment may include symptoms such as impaired muscle control, impaired movement, or other movement problems such as stiffness, spasticity, bradykinesia, rhythmic hyperkinesia, non-rhythmic hyperkinesia, and akinesia. In some cases, the movement disorder can be a symptom of parkinson's disease. However, dyskinesias can be attributed to other patient conditions.

In the example of fig. 1, system 100 is depicted as a DBS system. However, the techniques disclosed herein may be applied to other types of treatment systems for managing patient symptoms not explicitly shown in the example of fig. 1. For example, the disclosed techniques described herein may additionally be applied to systems that deliver Spinal Cord Stimulation (SCS) therapy for spinal cord injury or for the purpose of suppressing pain in the patient 112. Further, the techniques of the present disclosure may be applied to deliver pelvic stimulation (e.g., sacral neuromodulation) to systems delivering therapy for pelvic health and/or gastric applications.

Exemplary therapy system 100 includes a medical device programmer 104, an Implantable Medical Device (IMD)106, a lead extension 110, and leads 114A and 114B having respective electrode sets 116, 118. In the example shown in fig. 1, the electrodes 116, 118 of the leads 114A, 114B are positioned to deliver electrical stimulation to a tissue site within the brain 120, such as a deep brain site beneath the dura mater of the brain 120 of the patient 112. In some examples, delivering stimulation to one or more regions of brain 120, such as the subthalamic nucleus, the globus pallidus, or the thalamus, may be an effective treatment to manage movement disorders such as parkinson's disease. Some or all of the electrodes 116, 118 may also be positioned to sense bioelectrical signals within the brain 120 of the patient 112. In some examples, some of the electrodes 116, 118 may be configured to sense bioelectrical signals, and others of the electrodes 116, 118 may be configured to deliver adaptive electrical stimulation to the brain 120. In other examples, all of the electrodes 116, 118 are configured to sense bioelectrical signals and deliver adaptive electrical stimulation to the brain 120.

IMD106 includes a therapy module (e.g., which may include processing circuitry, signal generation circuitry, or other circuitry configured to perform functions attributed to IMD 106) that includes a stimulation generator configured to generate and deliver electrical stimulation therapy to patient 112 via a subset of electrodes 116, 118 of leads 114A and 114B, respectively. The subset of electrodes 116, 118 used to deliver electrical stimulation to the patient 112, and in some cases, the polarity of the subset of electrodes 116, 118, may be referred to as a stimulation electrode combination. As described in further detail below, stimulation electrode combinations may be selected (e.g., selected based on patient condition) for a particular patient 112 and target tissue site. The electrode sets 116, 118 include at least one electrode and may include a plurality of electrodes. In some examples, the plurality of electrodes 116 and/or 118 may have complex electrode geometries such that two or more electrodes are located at different positions around the perimeter of the respective lead.

In some examples, the bioelectrical signals sensed within brain 120 may reflect changes in current resulting from the summation of potential differences across brain tissue. Examples of bioelectric signals include, but are not limited to, electrical signals generated by Local Field Potentials (LFPs) sensed within one or more regions of the brain 120, such as electroencephalography (EEG) signals, corticoptogram (ECoG) signals, or other types of neurobrain signals. However, the local field potentials may include a wider variety of electrical signals within the brain 120 of the patient 112.

In some examples, the bioelectrical signals used to select the stimulation electrode combinations may be sensed within the same region of brain 120 as the target tissue site for electrical stimulation. As previously noted, these tissue sites may include tissue sites within the anatomy (such as the thalamus, subthalamic nucleus, or globus pallidus of the brain 120), as well as other target tissue sites. A particular target tissue site and/or region within brain 120 may be selected based on a patient condition. Thus, in some examples, both stimulation and sensing electrode combinations may be selected from the same set of electrodes 116, 118. In other examples, the electrodes used to deliver the electrical stimulation may be different than the electrodes used to sense the bioelectric signals.

The electrical stimulation generated by IMD106 may be configured to manage various disorders and conditions. In some examples, a stimulation generator of IMD106 is configured to generate and deliver electrical stimulation pulses to patient 112 via electrodes of the selected stimulation electrode combination. However, in other examples, the stimulation generator of IMD106 may be configured to generate and deliver a continuous wave signal, such as a sine wave or a triangle wave. In either case, a stimulation generator within IMD106 may generate electrical stimulation therapy for DBS according to the selected therapy program. In examples where IMD106 delivers electrical stimulation in the form of stimulation pulses, the therapy program may include a set of therapy parameter values (e.g., stimulation parameters), such as a stimulation electrode combination used to deliver stimulation to patient 112, a pulse frequency, a pulse width, and a current or voltage amplitude of the pulses. As previously indicated, the electrode combination may indicate the particular electrodes 116, 118 selected for delivering stimulation signals to the tissue of the patient 112, and the respective polarities of the selected electrodes.

IMD106 may be implanted within a subcutaneous pocket above the clavicle or, alternatively, on or within skull 122, within the abdomen of patient 112, or at any other suitable site within patient 112. Generally, IMD106 is constructed of biocompatible materials that resist erosion and degradation by body fluids. IMD106 may include a hermetic enclosure to substantially enclose components such as a processor, therapy module, and memory.

As shown in fig. 1, implant lead extension 110 is coupled to IMD106 via connector 108 (also referred to as a connector block or joint of IMD 106). In the example of fig. 1, lead extension 110 traverses from the implantation site of IMD106 and along the neck of patient 112 to skull 122 of patient 112 to enter brain 120. In the example shown in fig. 1, leads 114A and 114B (collectively, "leads 114") are implanted within the right and left brains, respectively, of patient 112 in order to deliver electrical stimulation to one or more regions of brain 120, which may be selected based on a condition or disorder of the patient being controlled by treatment system 100. However, a particular target tissue site and stimulation electrode for delivering stimulation to the target tissue site may be selected, for example, according to identified patient behavior and/or other sensed patient parameters. Other lead 114 and IMD106 implantation sites are contemplated. For example, in some examples, IMD106 may be implanted on or within skull 122. In some examples, lead 114 may be implanted within the same half-brain, or IMD106 may be coupled to a single lead implanted in a single half-brain.

Existing lead sets include axial leads carrying ring electrodes disposed at different axial locations and so-called "paddle" leads carrying planar array electrodes. The selection of electrode combinations within an axial lead, within a paddle lead, or between two or more different leads presents challenges to the clinician. In some examples, more or less complex lead array geometries and/or electrode array geometries may be used.

Although lead 114 is shown in fig. 1 as being coupled to common lead extension 110, in other examples, lead 114 may be coupled to IMD106 via a separate lead extension or directly to connector 108. Leads 114 may be positioned to deliver electrical stimulation to one or more target tissue sites within brain 120 to manage patient symptoms associated with dyskinesia of patient 112. The lead 114 may be implanted to position the electrodes 116, 118 at desired locations of the brain 120 through respective holes in the skull 122. The lead 114 may be placed at any location within the brain 120 such that the electrodes 116, 118 are capable of providing electrical stimulation to a target tissue site within the brain 120 during treatment. For example, electrodes 116, 118 may be surgically implanted beneath the dura mater of brain 120 or within the cerebral cortex of brain 120 via a bore in cranium 122 of patient 112 and electrically coupled to IMD106 via one or more leads 114. The lead 114 may also be placed at other locations within the central or peripheral nervous system as desired to sense or modulate nervous system activity. In other examples not depicted in the example of fig. 1, the lead 114 may be implanted in other locations within the patient 112, such as near the spinal cord, sacral nerves, or muscle fibers (e.g., for EMG).

In the example shown in fig. 1, the electrodes 116, 118 of the lead 114 are shown as ring electrodes. The ring electrode is useful in DBS applications because the ring electrode is relatively easy to program and is capable of delivering an electric field to any tissue adjacent to the electrodes 116, 118. In other examples, the electrodes 116, 118 may have different configurations. For example, at least some of the electrodes 116, 118 of the lead 114 may have a complex electrode array geometry capable of generating a shaped electric field. The complex electrode array geometry may include multiple electrodes (e.g., partial ring or segmented electrodes) around the outer periphery of each lead 114 instead of one ring electrode. In this way, electrical stimulation may be directed from lead 114 in a particular direction to enhance therapeutic efficacy and reduce possible adverse side effects due to stimulation of a large amount of tissue. In some examples, a housing of IMD106 may include one or more stimulation and/or sensing electrodes. In alternative examples, the lead 114 may have a shape other than an elongated cylinder as shown in fig. 1. For example, the lead 114 may be a paddle lead, a ball lead, a lead that is capable of bending, or any other type of shape effective in treating the patient 112 and/or minimizing the invasiveness of the lead 114.

IMD106 includes a memory for storing a plurality of therapy programs, each therapy program defining a set of therapy parameter values. In some examples, IMD106 may select a therapy program from memory based on various parameters (such as sensed patient parameters and identified patient behavior). IMD106 may generate electrical stimulation based on the selected therapy program to manage patient symptoms associated with the movement disorder. In some examples, the therapy program may be stored on another device (such as external programmer 104) or distributed on one or more computing devices (e.g., a cloud computing system).

External programmer 104 wirelessly communicates with IMD106 to provide or retrieve therapy information as needed. Programmer 104 is an external computing device that a user (e.g., a clinician and/or patient 112) may use to communicate with IMD 106. For example, programmer 104 may be a clinician programmer that a clinician uses to communicate with IMD106 and program IMD106 with one or more therapy programs. Alternatively, programmer 104 may be a patient programmer that allows patient 112 to select a program and/or view and modify treatment parameters. The clinician programmer may include more programming features than the patient programmer. In other words, only the clinician programmer may allow for more complex or sensitive tasks to prevent untrained patients from making undesirable changes to IMD 106.

When programmer 104 is configured for use by a clinician, programmer 104 may be used to transmit initial programming information to IMD 106. This initial information may include hardware information such as the type and electrode arrangement of lead 114, the location of lead 114 within brain 120, the configuration of electrode arrays 116, 118, the initial program defining the values of the therapy parameters, and any other information that the clinician desires to program into IMD 106. Programmer 104 is also capable of performing functional tests (e.g., measuring the impedance of electrodes 116, 118 of lead 114). Programmer 104 is also capable of downloading or streaming patient data from IMD106 and processing such patient data. In some examples, programmer 104 may download and process such patient data immediately or on a delayed or periodic basis. In other examples, programmer 104 may upload such patient data to one or more computing devices (e.g., a cloud computing network) for processing. In some examples, programmer 104 may upload such patient data immediately or on a delayed or periodic basis.

The clinician may also store a therapy program within IMD106 via programmer 104. During the programming session, the clinician may determine one or more therapy programs that may provide effective therapy to the patient 112 to address symptoms associated with the patient's condition, and in some cases, symptoms specific to one or more different patient states (such as a sleep state, a moving state, or a resting state). For example, a clinician may select one or more stimulation electrode combinations with which to deliver stimulation to brain 120. During the programming session, the clinician may evaluate the efficacy of a particular procedure evaluated based on feedback provided by the patient 112 or based on one or more physiological parameters of the patient 112 (e.g., one or more characteristics of one or more bioelectrical signals, muscle activity, muscle tension, stiffness, tremor, etc.). Alternatively, the identified patient behavior from the video information may be used as feedback during the initial programming session and subsequent programming sessions. Programmer 104 may assist a clinician in creating/identifying a treatment program by providing an organizational system for identifying potentially beneficial treatment parameter values.

The programmer 104 may also be configured for use by the patient 112. When configured as a patient programmer, programmer 104 may have limited functionality (as compared to a clinician programmer) in order to prevent patient 112 from altering critical functions of IMD106 or applications that may be harmful to patient 112. As such, programmer 104 may only allow patient 112 to adjust the values of certain treatment parameters or set the available range of values for particular treatment parameters.

Programmer 104 may also provide an indication to patient 112 when therapy is delivered, when patient input has triggered a therapy change, or when a power source within programmer 104 or IMD106 needs to be replaced or recharged. For example, programmer 104 may include an alert LED that may send a message to patient 112 via the programmer display, generating an audible sound or somatosensory prompt to confirm receipt of patient input, e.g., to indicate patient status or to manually modify a therapy parameter. As described in more detail below, in some examples, programmer 104 displays a notification to a clinician or patient that an artifact is present in one or more electrodes 116, 118 of lead 114.

The therapy system 100 may be implemented to provide chronic stimulation therapy to the patient 112 over the course of months or years. However, the system 100 may also be employed on a trial basis to evaluate treatment prior to full implantation. If implemented temporarily, some components of the system 100 may not be implanted in the patient 112. For example, patient 112 may be provided with an external medical device, such as a test stimulator, instead of IMD 106. The external medical device may be coupled to the percutaneous lead or the implanted lead via a percutaneous extension. If the trial stimulator instructs the DBS system 100 to provide effective therapy to the patient 112, the clinician may implant the chronic stimulator in the patient 112 for relatively long-term therapy.

Although IMD 104 is described as delivering electrical stimulation therapy to brain 120, IMD106 may be configured to direct electrical stimulation to other anatomical regions of patient 112. In other examples, system 100 may include an implantable drug pump in addition to IMD106 or instead of IMD 106. In addition, IMDs may provide other electrical stimulation, such as spinal cord stimulation, to treat movement disorders.

In accordance with the techniques of this disclosure, the system 100 may sense bioelectrical signals of the brain 120 of the patient 112 via the electrodes 116, 118 and determine a probability entropy of the bioelectrical signals. The system 100 may use the probability entropy of the bioelectrical signals to distinguish between ones of the electrodes 116, 118 that are capable of sensing clean bioelectrical signals and ones of the electrodes 116, 118 that are contaminated by artifacts. Bioelectrical signals, such as neuronal LFP activity, in the brain 120 of the patient 112 may generally exhibit random (e.g., stochastic) behavior and exhibit high entropy. In contrast, the patterns or orders in the sensed bioelectrical signal exhibit low entropy. Low entropy in the sensed signal may indicate artifacts in the sensed signal, such as periodic artifacts that occur due to ECG, motion, or other periodic noise sources. In other words, if the sensed signal exhibits random behavior (e.g., a random process), and thus exhibits high entropy, the sensed signal tends not to include artifacts. However, if the sensed signal exhibits periodic components (e.g., patterns or high orders), and thus exhibits low entropy, the sensed signal may include artifacts.

In some examples, the system 100 derives a probability entropy value for the bioelectric signal from a probability distribution of values of the bioelectric signal over a period of time. For example, B [ i ] may represent a histogram of the measurement of the value of the bioelectrical signal over a period of time, and f [ i ] may represent a fraction of the value of the bioelectrical signal in B [ i ] (e.g., the number of "slices" or "segments" of the histogram B [ i ]). This can be calculated, for example, using shannon entropy from a histogram of values. For example, let f [ i ] equal a fraction of the values in b [ i ]. The shannon entropy is then defined by the following equation:

for f [ i ]]＞0，Shannon Entropy＝-sum(f[i]*log₂(f[i]))

A uniform distribution of values indicates high entropy in the sensed bioelectric signal, such as may be the case for a white noise signal. In contrast, an inhomogeneous distribution of values indicates low entropy in the sensed bioelectric signal, such as may be the case for signals comprising periodic or sinusoidal components. The techniques of this disclosure recognize that low-entropy signals that include periodic or sinusoidal components may include artifacts that contaminate the signal, while high-entropy signals may not include such periodic or sinusoidal components, and may more accurately represent random, high-entropy bioelectrical signals of brain 120 of patient 112.

For example, an uneven distribution may occur for sensed bioelectrical signals comprising periodic or sinusoidal components. Because the probability distribution of the values of the periodic signal exhibits a constrained frequency spectrum, the spectral power of the periodic signal may exhibit a higher order and lower entropy than, for example, a white noise signal (e.g., a signal that is random or exhibits a high entropy). The periodic signal has a constrained spectrum and therefore, based on spectral power, the entropy of the periodic signal is lower than that of, for example, a white noise spectrum. In this example, the entropy value of the periodic signal may be relatively low because of the order or pattern present in the signal. In some cases, the strong pattern may indicate the presence of a desired signal (e.g., a desired component of a sensed signal, such as an LFP record from the patient 112). In other cases, such strong periodicity may indicate the presence of an undesired signal source (e.g., a component of the sensed signal due to undesired noise). For example, if the sensed signal includes a component that exhibits strong periodicity in a spectrum outside of the spectrum of biomarkers of neuronal activity, the system 100 may classify the sensed signal as contaminated by artifacts.

In contrast, where the bioelectric signals exhibit a high entropy, a uniform or broad distribution of values may occur. For example, if the probability distribution values of the spectral power spread relatively evenly (e.g., none of the components of the sensed signal is more likely than any other value), as is the case for many bioelectrical signals, the sensed signal may not have a dominant periodic signal component. In this example, the entropy value may be relatively high because there are lower orders or patterns in the sensed signal. Further, the system 100 may classify signals having such high entropy values as "clean," e.g., exhibiting little or no artifacts.

In some examples, the probability entropy value of the bioelectric signal is a statistical measure of the randomness of the value of the bioelectric signal over a period of time. Thus, the probability entropy value of the bioelectric signal is a measure of the degree of randomness of the bioelectric signal. In some examples, system 100 may verify that the bioelectric signals sensed by one of electrodes 116, 118 are of sufficient quality to be used as biomarkers for controlling the aDBS therapy. In some examples, the system 100 may use the probability entropy of the bioelectrical signal sensed by one of the recording electrodes 116, 118 as an indicator of interference from other periodic bioelectrical signals (such as ECG or other types of periodic artifacts present in the recorded bioelectrical signal of the brain 120 of the patient 112).

In one example, IMD106 senses a plurality of bioelectrical signals of brain 120 of patient 112 via electrodes 116, 118. IMD106 determines a probability entropy value for the bioelectrical signals for each bioelectrical signal sensed at the respective electrode 116, 118. IMD106 compares each of the respective probability entropy values of the sensed bioelectrical signals to a respective entropy threshold. IMD106 selects a subset of electrodes 116, 118 based on the comparison. IMD106 controls delivery of electrical stimulation therapy to patient 112 based on the bioelectrical signals sensed via respective electrodes of the subset of electrodes 116, 118 and excludes bioelectrical signals of the plurality of bioelectrical signals that are sensed via respective electrodes 116, 118 that are not in the subset.

For example, as described above, in the aDBS, IMD106 may determine a therapy to deliver based at least in part on the sensed bioelectrical signals. If the sensed signal has low entropy, there is a possibility that an artifact exists in the sensed signal, and IMD106 may not be able to fully distinguish between the actual bioelectric signal and the artifact. In such examples, IMD106 may not rely on sensed signals with artifacts to determine the therapy to deliver. If the sensed signal has high entropy, there is a likelihood that the sensed signal is an accurate representation of the patient-generated bioelectric signal, and IMD106 may rely on the sensed signal to determine the therapy to be delivered.

In the foregoing example, the above-described techniques for distinguishing between electrodes capable of sensing clean bioelectric signals and electrodes contaminated by artifacts are performed by IMD 106. However, in other examples of the techniques of this disclosure, may be performed by external programmer 104. In further examples, the techniques of this disclosure are performed by one or more computing devices (such as laptops, tablets, smartphones, PDAs, cloud computing systems, etc.) in communication with IMD106, not depicted in fig. 1. In still other examples, the techniques of this disclosure may be performed by IMD106, external programmer 104, or a combination of any one or more of the foregoing computing devices.

Accordingly, the techniques disclosed herein may provide enhanced accuracy in the identification of artifacts in the electrodes. For example, the techniques of this disclosure may detect artifacts that may otherwise be difficult to detect using conventional artifact detection methods, such as artifacts with variable signal amplitudes across multiple recordings. Thus, by identifying and eliminating measurements from artifact-contaminated electrodes, the techniques of the present disclosure may provide greater reliability in that the signals sensed by the electrodes and used as biomarkers for the aDBS accurately reflect true bioelectric signals and avoid erroneous measurements that may adversely affect the therapy provided to the patient. Accordingly, the techniques disclosed herein may provide patients with more effective aDBS therapy than conventional systems.

Fig. 2 is a block diagram of the exemplary IMD106 of fig. 1 for delivering adaptive deep brain stimulation therapy. In the example shown in fig. 2, IMD106 includes processing circuitry 210, memory 211, stimulation generation circuitry 202, sensing circuitry 204, switching circuitry 206, telemetry circuitry 208, sensors 212, and power supply 220. Each of these circuits may be or include a circuit configured to perform a function attributed to each respective circuit. For example, the processing circuitry 210 may include processing circuitry, the switching circuitry 206 may include switching circuitry, the sensing circuitry 204 may include sensing circuitry, and the telemetry circuitry 208 may include telemetry circuitry. The memory 211 may include any volatile or non-volatile media, such as Random Access Memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), Electrically Erasable Programmable ROM (EEPROM), flash memory, and the like. Memory 211 may store computer readable instructions that, when executed by processing circuitry 210, cause IMD106 to perform various functions. The memory 211 may be a storage device or other non-transitory medium.

In the example shown in fig. 2, memory 211 stores therapy programs 214 and sensing electrode combinations and associated stimulation electrode combinations 218 in separate memories within memory 211 or in separate areas within memory 211. Each stored therapy program 214 defines a particular set of electrical stimulation parameters (e.g., a set of therapy parameters), such as stimulation electrode combinations, electrode polarities, current or voltage amplitudes, pulse widths, and pulse frequencies. In some examples, the individual therapy programs may be stored as a therapy group that defines a set of therapy programs that may be used to generate stimulation. The stimulation signals defined by the therapy programs of the therapy groups may be delivered together on an overlapping or non-overlapping (e.g., time-staggered) basis.

The sensing and stimulation electrode combinations 218 store the sensing electrode combinations and associated stimulation electrode combinations. As described above, in some examples, sensing and stimulation electrode combination 218 may include the same subset of electrodes 116, 118, a housing of IMD106 that serves as an electrode, or may include a different subset or combination of such electrodes. Thus, memory 211 may store a plurality of sensing electrode combinations and, for each sensing electrode combination, information identifying the stimulation electrode combination associated with the respective sensing electrode combination. The association between the sensing electrode combination and the stimulation electrode combination may be determined, for example, by a clinician or automatically by the processing circuit 210. In some examples, the corresponding sensing and stimulation electrode combinations may include some or all of the same electrodes. However, in other examples, some or all of the electrodes in the corresponding sensing and stimulation electrode combinations may be different. For example, a stimulation electrode combination may include more electrodes than a corresponding sensing electrode combination in order to increase the efficacy of the stimulation therapy. In some examples, as discussed above, stimulation may be delivered to a tissue site that is different from the tissue site closest to the corresponding sensing electrode combination but within the same region of brain 120 (e.g., the thalamus) via a stimulation electrode combination in order to mitigate any erratic oscillations or other irregular brain activity within the tissue site associated with the sensing electrode combination.

Under control of the processing circuitry 210, the stimulation generation circuitry 202 generates stimulation signals for delivery to the patient 112 via the selected combination of electrodes 116, 118. Example ranges of electrical stimulation parameters believed to be effective in DBS to manage dyskinesia in patients include:

1. pulse frequency, i.e., frequency: between about 40 hertz and about 500 hertz, such as between about 90 hertz and 170 hertz or such as about 90 hertz.

2. For the voltage control system, the voltage amplitude: between about 0.1 volts and about 50 volts, such as between about 2 volts and about 3 volts.

3. In the alternative case of the current control system, the current amplitude: between about 1 milliamp and about 3.5 milliamps, such as between about 1.0 milliamp and about 1.75 milliamps.

4. Pulse width: between about 50 microseconds and about 500 microseconds, such as between about 50 microseconds and about 200 microseconds.

Thus, in some examples, stimulation generation circuitry 202 generates electrical stimulation signals according to the electrical stimulation parameters described above. Other ranges of treatment parameter values may also be useful and may depend on the target stimulation site within the patient 112. Although stimulation pulses are described, the stimulation signals may be in any form, such as continuous time signals (e.g., sine waves), and the like.

The processing circuitry 210 may include fixed function processing circuitry and/or programmable processing circuitry, and may include, for example, one or more of the following: a microprocessor, a controller, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), a discrete logic circuit, or any other processing circuit configured to provide the functionality attributed to processing circuit 210, processing circuit 210 may be embodied herein as firmware, hardware, software, or any combination thereof. Processing circuitry 210 may control stimulation generation circuitry 202 in accordance with therapy programs 214 stored in memory 211 to apply particular stimulation parameter values, such as voltage amplitude or current amplitude, pulse width, or pulse frequency, specified by one or more of the programs.

In the example shown in fig. 2, the set of electrodes 116 includes electrodes 116A, 116B, 116C, and 116D, and the set of electrodes 118 includes electrodes 118A, 118B, 118C, and 118D. The processing circuit 210 also controls the switching circuit 206 to apply the stimulation signals generated by the stimulation generation circuit 202 to selected combinations of electrodes 116, 118. In particular, the switching module 204 may couple the stimulation signal to selected conductors within the lead 114, which in turn delivers the stimulation signal across the selected electrodes 116, 118. The switching circuit 206 may be a switch array, switch matrix, multiplexer, or any other type of switching module configured to selectively couple stimulation energy to selected electrodes 116, 118 and to selectively sense a bioelectric signal with the selected electrodes 116, 118. Accordingly, stimulation generation circuitry 202 is coupled to electrodes 116, 118 via conductors within switching circuitry 206 and lead 114. However, in some examples, IMD106 does not include switching circuitry 206.

The stimulus generation circuit 202 may be a single channel or multi-channel stimulus generator. In particular, stimulation generation circuitry 202 may be capable of delivering a single stimulation pulse, multiple stimulation pulses, or a continuous signal at a given time via a single electrode combination, or multiple stimulation pulses at a given time via multiple electrode combinations. However, in some examples, the stimulus generation circuit 202 and the switching circuit 206 may be configured to deliver multiple channels on a time-interleaved basis. For example, the switching circuitry 206 may be used to time-divide the output of the stimulation generation circuitry 202 on different electrode combinations at different times to deliver multiple programs or channels of stimulation energy to the patient 112. Alternatively, the stimulus generation circuit 202 may include a plurality of voltage or current sources and receivers coupled to respective electrodes to drive the electrodes as cathodes or anodes. In this example, IMD106 may not require the time-interleaved multiplexing functionality of switching circuitry 206 to stimulate via different electrodes.

The electrodes 116, 118 on the respective leads 114 may be constructed of a number of different designs. For example, one or both of leads 114 may include two or more electrodes at each longitudinal position along the length of the lead, such as a plurality of electrodes at different circumferential positions around the circumference of the lead at each of positions A, B, C and D. In one example, the electrodes may be electrically coupled to the switching circuit 206 via respective wires that are straight or coiled within the housing of the lead and extend to a connector at the proximal end of the lead. In another example, each of the electrodes of the lead may be an electrode deposited on the thin film. The film may include a conductive trace for each electrode that extends along the length of the film to a proximal end connector. The film may then be wrapped (e.g., spirally wrapped) around the inner member to form the lead 114. These and other configurations may be used to form leads with complex electrode geometries.

Although sensing circuitry 204 is incorporated into a common housing with stimulation generation circuitry 202 and processing circuitry 210 in fig. 2, in other examples, sensing circuitry 204 may be located in a housing separate from IMD106 and may communicate with processing circuitry 210 via wired or wireless communication techniques. Exemplary bioelectric signals include, but are not limited to, signals generated by Local Field Potentials (LFPs) within one or more regions of the brain 28. EEG and ECoG signals are examples of local field potentials that may be measured within brain 120. However, the local field potentials may include a wider variety of electrical or neural signals within the brain 120 of the patient 112. Other examples of bioelectric signals may include signals from nerve fibers (e.g., spinal cord) of a patient sensed via EEG or ECoG or signals from muscle fibers of a patient sensed via EMG.

Telemetry circuitry 208, under the control of processing circuitry 210, supports wireless communication between IMD106 and external programmer 104 or another computing device. As an update to the program, processing circuitry 210 of IMD106 may receive values of various stimulation parameters, such as magnitude and electrode combinations, from programmer 104 via telemetry circuitry 208. Updates to the treatment program may be stored in the treatment program 214 portion of the memory 211. Telemetry circuitry 208 in IMD106 and telemetry modules in other devices and systems described herein, such as programmer 104, may communicate via Radio Frequency (RF) communication techniques. Further, telemetry circuitry 208 may communicate with external medical device programmer 104 via proximal inductive interaction of IMD106 with programmer 104. Thus, telemetry circuitry 208 may transmit information to external programmer 104 continuously, at periodic intervals, or upon request from IMD106 or programmer 104.

Power supply 220 delivers operating power to various components of IMD 106. Power supply 220 may include a small rechargeable or non-rechargeable battery and power generation circuitry to generate operating power. Recharging may be accomplished by proximal inductive interaction between an external charger and an inductive charging coil or other power transmission mechanism or modality within IMD 220. In some examples, the power requirements may be small enough to allow IMD 220 to take advantage of patient motion and implement a kinetic energy scavenging device to trickle charge the rechargeable battery. In other examples, a conventional battery may be used for a limited period of time.

In one example, processing circuitry 210 of IMD106 senses one or more bioelectrical signals of brain 120 of patient 112 via electrodes 116, 118 (and sensing circuitry 202) interposed along leads 114. Further, processing circuitry 210 of IMD106 delivers electrical stimulation therapy to patient 112 via electrodes 116, 118 (and stimulation generation circuitry 202) based on the sensed one or more bioelectrical signals of brain 120. The adaptive DBS therapy is defined by one or more therapy programs 214 having one or more parameters stored in memory 211. For example, the one or more parameters include current amplitude (for a current control system) or voltage amplitude (for a voltage control system), pulse frequency or frequency, and pulse width, or number of pulses per cycle. In examples where electrical stimulation is delivered according to "bursts" of pulses or a series of electrical pulses defined by "on-times" and "off-times," the one or more parameters may also define one or more of a number of pulses per burst, an on-time, and an off-time. Processing circuitry 210 delivers adaptive DBS to patient 112 via electrodes 116, 118, and may adjust one or more parameters defining electrical stimulation based on corresponding parameters of the sensed one or more bioelectrical signals of brain 120.

In some examples, the processing circuitry 210 measures the one or more bioelectrical signals continuously in real-time. In other examples, the processing circuit 210 samples the one or more bioelectrical signals according to a predetermined frequency or periodically after a predetermined amount of time. In some examples, the processing circuit 210 periodically samples the signal at a frequency of about 200 hertz. In some examples, the processing circuit 210 periodically samples the signal at a frequency of about 250 hertz.

In accordance with techniques of this disclosure, IMD106 may use probabilistic entropy of one or more bioelectrical signals of brain 120 of patient 112 sensed via electrodes 116, 118 to distinguish between electrodes capable of sensing clean bioelectrical signals and electrodes contaminated with artifacts. The techniques of this disclosure recognize that bioelectrical signals, such as neuronal LFP activity, in the brain 120 of the patient 112 may generally exhibit random (e.g., stochastic) behavior and exhibit high entropy in the spectral components of the recorded signals. In contrast, the techniques of this disclosure recognize that patterns or orders in the band power of the sensed bioelectric signal (e.g., low entropy in the band power) may indicate artifacts in the sensed signal, such as artifacts that occur due to ECG, motion, or other types of periodic artifacts. In some examples, IMD106 may verify that the bioelectrical signals sensed by one of electrodes 116, 118 are of sufficient quality to be used as biomarkers to control therapy delivery (such as an adss therapy). In some examples, IMD106 may use the probability entropy of the bioelectrical signals sensed by each of recording electrodes 116, 118 as an indicator of ECG or other periodic artifacts present in the recorded LFP signals of brain 120 of patient 112 for that particular one of electrodes 116, 118.

In one example, the processing circuitry 210 senses a plurality of bioelectrical signals of the brain 120 of the patient 112 via the electrodes 116, 118 and the sensing circuitry 204. The processing circuitry 210 determines a probability entropy value for the bioelectrical signals for each bioelectrical signal sensed at the respective electrode 116, 118. Additional examples of how the processing circuit 210 may determine probability entropy are provided in more detail below. Processing circuitry 210 compares each of the respective probability entropy values of the sensed bioelectrical signals to a respective entropy threshold.

In some examples, the entropy threshold is defined by a clinician. In some examples, the entropy threshold is generated by a machine learning model. For example, a machine learning model may be trained with training data that includes a plurality of bioelectrical signals from a plurality of patients, each bioelectrical signal labeled with data indicating whether the bioelectrical signal includes one or more artifacts and the location of the one or more artifacts (if present). In some examples, the machine learning model is a supervised learning algorithm that uses training data that includes input features with associated target labels. In some examples, the machine learning model is a logistic regression, a support vector machine, a random forest, or a gradient elevator. In some examples, the machine learning model receives as input features one or more features of the sensed bioelectrical signals (such as entropy with power, threshold crossing rate, and/or entropy of inter-threshold crossing intervals). In some examples, the input is tagged with an object label that defines a portion of the signal that exhibits or is absent of an artifact. The machine learning model may process the training data to determine a relationship of one or more characteristics of one or more features of the bioelectrical signal to the presence of artifacts in the bioelectrical signal. For example, the machine learning model may determine a correlation of one or more characteristics of one or more features of the bioelectrical signal with the presence of artifacts in the bioelectrical signal, and a strength of the correlation. In one example, the machine learning model uses one or more of one or more characteristics of one or more features of the bioelectrical signal, a correlation of the one or more characteristics with the presence of artifacts in the bioelectrical signal, or a strength of the correlation as an entropy threshold.

Further, the machine learning model may iteratively process the training data to determine different weights corresponding to the strength of the correlation of the one or more characteristics of the one or more features of the bioelectrical signal with the presence of the artifact in the bioelectrical signal. In examples described below that use multiple methods to determine the presence of artifacts in bioelectrical signals, the machine learning model may apply weights to each respective method in order to more accurately identify the presence of artifacts in bioelectrical signals.

The processing circuitry 210 may determine the probability entropy value of the bioelectric signal in a variety of ways. As an example, which will be described in more detail below, the processing circuitry 210 may determine the probability entropy value of the bioelectric signal by: 1) analyzing a statistical measure of randomness of spectral power over a plurality of frequency bands of the bioelectric signal; or 2) comparing the entropy of one or more features of the bioelectrical signal to a threshold limit. However, processing circuitry 210 may determine the probability entropy value of the bioelectrical signal using other methods contemplated by techniques of the present disclosure not explicitly described herein.

The probability entropy value is determined by analyzing a statistical measure of randomness of spectral power over a plurality of frequency bands of the bioelectric signal.

As one example, the processing circuitry 210 may determine the probability entropy value of the bioelectric signal by analyzing a statistical measure of randomness of spectral power over a plurality of frequency bands of the bioelectric signal. The processing circuit 210 splits the bioelectric signal into a plurality of frequency bands. In some examples, the processing circuitry 210 applies a Welch method to determine the spectral power of each frequency band of the bioelectric signal. In one example, the processing circuit 210 samples the LFP signal in the time domain at a frequency of 250 hertz and divides the sampled LFP signal into 1 second segments having 50% overlap. For example, a first segment includes values of the sampled LFP signal from 0 to 1 second, a second segment includes values of the sampled LFP signal from 0.5 to 1.5 seconds, a third segment includes values of the sampled LFP signal from 1 to 2 seconds, a fourth segment includes values of the sampled LFP signal from 1.5 to 2.5 seconds, and so on. Processing circuit 210 applies a hanning window to window each segment of the sampled LFP signal. The processing circuit 210 computes a periodogram for each 1-second segment of the LFP signal that depicts the Power Spectral Density (PSD) over the 1-second segment. In some examples, the periodogram depicts a square of a magnitude of a Fast Fourier Transform (FFT) for each windowed 1-second segment of the sampled LFP signal. The processing circuit 210 averages the periodograms to calculate the power spectral density of the sampled LFP signal for each frequency band. Thus, the processing circuit 210 may use the periodogram to empirically determine a statistical distribution of spectral power over each frequency band, and may average the periodogram to produce an average power spectral density, i.e., a distribution of power per spectral band.

The processing circuit 210 determines a statistical measure of the randomness of the calculated power spectral density over the frequency band of the sampled LFP signal. As one example, if the calculated power spectral density has an uneven distribution of values, the sampled LFP signal exhibits low entropy and may include periodic or sinusoidal components. Thus, in this example, an uneven distribution of values in the power spectral density of the sampled LFP signal may indicate that the sampled LFP signal includes artifacts that contaminate the signal. For example, the processing circuit 210 compares the calculated statistical measure of randomness of the power spectral density to a corresponding entropy threshold. If the statistical measure of randomness of the calculated power spectral densities is less than an entropy threshold (which may be the case if a particular band of the sampled LFP signal has a periodic or sinusoidal component), the statistical measure of randomness is indicative of the bioelectric signal contaminated by artifacts.

In another example, the sampled LFP signal exhibits high entropy if the calculated power spectral density has a uniform distribution of values. Thus, in this example, a uniform distribution of values in the power spectral density of the sampled LFP signal may indicate that the sampled LFP signal has no or little artifacts. For example, the processing circuit 210 compares the calculated statistical measure of randomness of the power spectral density to a corresponding entropy threshold. If the statistical measure of the randomness of the calculated power spectral densities is larger than an entropy threshold, which may be the case if a particular frequency band of the sampled LFP signal does not exhibit a larger power density than any other frequency band of the sampled LFP signal, the statistical measure of randomness indicates a bioelectric signal that is not or hardly contaminated by artifacts.

In the above example, the processing circuit 210 determines a statistical measure of the randomness of the calculated power spectral density over all bands of the sampled LFP signal to quantify the entropy of the entire sampled LFP signal. However, in other examples, the processing circuit 210 may determine a statistical measure of the randomness of the calculated power spectral densities for only a portion of the sampled LFP signals. For example, the processing circuit 210 may determine a first statistical measure of the randomness of the calculated power spectral densities of a first frequency band of the sampled LFP signals and determine a second statistical measure of the randomness of the calculated power spectral densities of a second frequency band of the sampled LFP signals. In one example, the processing circuit 210 determines that the sampled LFP signal is contaminated by artifacts if a first statistical measure of randomness of the power spectral density of the first frequency band indicates that the sampled LFP signal is contaminated by artifacts, but a second statistical measure of randomness of the power spectral density of the second frequency band indicates that the sampled LFP signal is clean. In another example, the processing circuit 210 determines that the sampled LFP signal is contaminated by artifacts only if the first and second statistical measures of randomness of the power spectral densities of the first and second frequency bands both indicate that the sampled LFP signal is contaminated by artifacts.

The probability entropy value is determined by comparing the entropy of one or more features of the bioelectrical signal to a threshold limit.

As another example, the processing circuitry 210 may determine the probability entropy value of the bioelectric signal by comparing the entropy of one or more features of the bioelectric signal to a threshold limit. For example, the processing circuit 210 uses a classification method to compare multiple features of the bioelectrical signal (such as spectral entropy, threshold crossing rate values, and/or inter-threshold crossing interval times) to one another. For example, such a classifier may classify artifact and/or normal signal type, return a probability of artifact in the bioelectric signal, or provide another rating of signal quality with respect to signal regularity or lack of signal regularity. In this example, features that exceed a threshold limit compared to other features of the sensed bioelectrical signal may be identified as statistically abnormal values. The processing circuit 210 may determine that such statistical outliers indicate that the bioelectric signal may be contaminated by artifacts in several ways, which will be described in more detail below.

As a first example, the processing circuitry 210 may determine the probability entropy value of the bioelectrical signal by analyzing a rate over time at which a characteristic of the bioelectrical signal exceeds a threshold limit. As an example of the one or more characteristics of the bioelectric signal being the amplitude of the bioelectric signal, the processing circuitry 210 may determine the probability entropy value of the bioelectric signal by analyzing a rate over time at which the amplitude of the bioelectric signal exceeds a threshold limit. If a large number of such amplitude threshold crossings occur within a particular time period, the bioelectric signal may be contaminated by artifacts.

For example, the processing circuit 210 compares the determined rate at which the amplitude exceeds a threshold limit to a rate threshold of the bioelectric signal. As one example, the processing circuit 210 determines a rate over time at which the amplitude of the first bioelectric signal exceeds a threshold limit and a rate over time at which the amplitude of the second bioelectric signal exceeds a threshold limit. If the rate of the first bioelectric signal is less than the rate threshold, the rate of the first bioelectric signal is indicative of a bioelectric signal with no or little artifact. Further, if the rate of the second bioelectric signal is greater than the rate threshold, the rate of the second bioelectric signal is indicative of a bioelectric signal contaminated by the artifact. In some examples, the rate over time at which the amplitude of the bioelectric signal exceeds a threshold limit may be used as an input to a machine learning system that processes one or more inputs to determine whether the input indicates the presence of an artifact in the bioelectric signal, as described above.

In some examples, the threshold limit is a first quartile range of the sensed amplitude of the bioelectric signal. In some examples, the processing circuit 210 determines the threshold limit as a normalized quartile range of the bioelectric signal minus a median value. For example, the processing circuit 210 may determine a median-subtracted normalization of the bioelectric signal by calculating a median of the sampled amplitudes of the bioelectric signal and subtracting the median from the sampled amplitudes of the bioelectric signal, and then dividing the resulting value by the interquartile range.

As a second example, the processing circuitry 210 may determine the probability entropy value for the bioelectrical signal by analyzing entropy of lengths of consecutive intervals between instances where one or more characteristics of the bioelectrical signal exceed a signal threshold limit. As an example of the one or more characteristics of the bioelectric signal being the amplitude of the bioelectric signal, the processing circuit 210 determines a time interval between a first instance in which the amplitude of the bioelectric signal exceeds a signal threshold limit and a second instance in which the amplitude of the bioelectric signal exceeds the signal threshold limit. In some examples, the signal threshold limit is a first quartile range of the sensed amplitude of the bioelectric signal. In some examples, processing circuitry 210 determines shannon entropy for a set of time intervals. For example, the processing circuit 210 may determine a time series of inter-threshold crossing intervals, as described above. Processing circuit 210 constructs a histogram of a plurality of time intervals, wherein each time interval is between two instances where the amplitude of the bioelectric signal exceeds a signal threshold limit.

The processing circuitry 210 may determine a probability entropy value for the bioelectric signal based on the resulting entropy exhibited by the histograms for the plurality of time intervals. For example, the processing circuit 210 compares the entropy of the time interval to an entropy threshold of the bioelectric signal. The techniques of this disclosure recognize that as entropy increases over a time interval between two consecutive instances in which the amplitude of the bioelectric signal exceeds a threshold limit, the likelihood of the presence of artifacts in the signal decreases. Thus, if the entropy of such detected intervals is less than the entropy threshold, the entropy indicates a bioelectric signal that may contain artifacts. Conversely, if the entropy of such detected intervals is greater than the entropy threshold, the entropy indicates a bioelectric signal with little or no artifacts. In some examples, the processing circuit 210 separately performs such analysis of the entropy of the time interval between two consecutive threshold crossings for a plurality of different threshold limits (e.g., a first quartile range of the time interval between two instances where the amplitude of the bioelectric signal exceeds the threshold limit, a second quartile range of the time interval between two instances where the amplitude of the bioelectric signal exceeds the threshold limit, etc.).

As a specific example, bioelectric signals containing artifacts of periodic waveforms (such as sine waves) may cross threshold limits at fairly regular intervals. Thus, the entropy decreases in the time interval between two consecutive signal crossings of the threshold limit. This reduction in signal cross entropy indicates that artifacts may be present.

In some examples, the entropy of the interval between a first instance of the amplitude of the bioelectric signal exceeding a signal threshold limit and a second instance of the amplitude of the bioelectric signal exceeding the signal threshold limit may be used as an input to a machine learning system that processes one or more inputs to determine whether the input indicates the presence of an artifact in the bioelectric signal, as described above. In some examples, if the processing circuit 210 identifies that no occurrence of the amplitude of the bioelectric signal exceeding the threshold limit occurs, there is no time interval between a first instance in which the amplitude of the bioelectric signal exceeds the threshold limit and a second instance in which the amplitude of the bioelectric signal exceeds the threshold limit. In such a case, the likelihood of artifacts in the signal may be very low. Accordingly, the processing circuitry 210 may assign a value of "-1" to the probability entropy value of the bioelectrical signal to push the determination of the machine learning system towards determining that the bioelectrical signal does not include artifacts.

A combination of techniques is used to determine probability entropy values.

In some examples, the processing circuitry 210 may determine the probability entropy value for the bioelectric signal by combining a plurality of techniques for determining the probability entropy value described above. For example, the processing circuitry 210 may apply two or more of the techniques described above to one or more features of the sensed bioelectrical signal, and apply different weights to each result to generate a probability entropy value for the bioelectrical signal that is more accurate than using a single method alone.

For example, the processing circuitry 210 may determine the probability entropy value of the bioelectrical signal using a combination of any of: (1) analyzing a statistical measure of randomness of spectral power over a plurality of frequency bands of the bioelectric signal; (2) analyzing a rate over time at which an amplitude of the bioelectrical signal exceeds a signal threshold limit; or (3)) analyzing the entropy of the time interval between two consecutive instances where the amplitude of the bioelectrical signal exceeds a signal threshold limit. In such examples, the processing circuitry 210 may determine that the bioelectric signal is contaminated by the artifact if the artifact is indicated by only one of the techniques described above, by more than one of the techniques described above, or by all of the techniques described above. In some examples, different weights may be applied to each of the above techniques, each weight corresponding to the strength of the correlation of that technique with the presence of artifacts in the bioelectric signal.

In some examples, prior to determining the probability entropy value of the bioelectric signal, the processing circuitry 210 may determine a normalization of the bioelectric signal. In this example, the processing circuitry 210 may then determine a probability entropy value for the normalized bioelectric signal using the techniques described above. By determining a normalization of the bioelectrical signal, the processing circuitry 210 may allow for a unified and simplified analysis of a plurality of bioelectrical signals having a plurality of different characteristics and waveforms.

The processing circuit 210 selects a subset of the electrodes 116, 118 based on the comparison. In some examples, the subset of electrodes 116, 118 includes only those electrodes whose corresponding bioelectric signals are determined by the processing circuitry 210 to indicate no or minimal artifacts. In other words, its corresponding bioelectric signal is determined by the processing circuitry 210 to indicate that the electrode with the artifact is excluded from the subset. The processing circuitry 210 controls delivery of the electrical stimulation therapy to the patient 112 based on the bioelectrical signals sensed via respective electrodes of the subset of electrodes 116, 118 and excludes bioelectrical signals of the plurality of bioelectrical signals that are sensed via respective electrodes 116, 118 that are not in the subset.

In some examples, the processing circuit 210 executes a machine learning system trained as described above. In some examples, the machine learning system implements a logistic regression model, a Support Vector Machine (SVM) model, a regression tree, a gradient boosting model, or other type of neural network model to create an artifact detector that uses probability entropy values of the bioelectric signals determined by the processing circuit 210 as inputs. For example, the machine learning system may process the respective probability entropy values of the bioelectrical signals determined by the processing circuitry 210 to identify one or more electrodes that meet the criteria of the machine learning model trained using the process described above. In some examples, the identified one or more electrodes are electrodes that satisfy the following condition: for the electrodes, the bioelectrical signals sensed via the identified one or more electrodes show no or little artifact. In some examples, the machine learning system processes probability entropy values of various different types of bioelectrical signals to identify one or more electrodes. For example, the machine learning system may process statistical measures of randomness of spectral power over multiple frequency bands of the respective bioelectrical signals sensed via the electrodes, a rate at which amplitudes of the respective bioelectrical signals sensed via the electrodes exceed an amplitude threshold limit, and an entropy of time intervals between successive instances at which amplitudes of the bioelectrical signals sensed via the electrodes exceed the threshold limit to identify one or more electrodes. In some examples, the machine learning system may assign different weights or coefficients to different types of probability entropy values for the bioelectrical signals determined during the training process described above. For example, the machine learning system may determine, based on the machine learning model, that a statistical measure of randomness of spectral power over multiple frequency bands of the bioelectrical signal should be given less weight than the rate at which the amplitude of the bioelectrical signal exceeds an amplitude threshold limit. As another example, the machine learning system may determine, based on the machine learning model, that a statistical measure of randomness of spectral power over multiple frequency bands of the bioelectrical signal should be given more weight than the entropy of the time interval between consecutive instances where the amplitude of the bioelectrical signal exceeds a threshold limit. The processing circuitry 210 selects the identified electrodes as a subset of the electrodes 116, 118 for controlling delivery of therapy to the patient 112.

In some examples, processing circuitry 210 transmits, via telemetry circuitry 208, an indication that an artifact is present in electrodes excluded from the subset. In some examples, the indication causes programmer 104 to display a notification to a clinician or patient that artifacts are present in electrodes excluded from the subset.

Accordingly, the techniques disclosed herein may provide enhanced accuracy in the identification of artifacts in the electrodes. For example, the techniques of this disclosure may detect artifacts that may otherwise be difficult to detect using conventional artifact detection methods, such as artifacts with variable signal amplitudes across multiple recordings. Thus, by identifying and eliminating measurements from artifact-contaminated electrodes, the techniques of this disclosure may provide greater reliability in an acdb system. For example, the techniques of this disclosure may increase the following possibilities: the signals sensed by the electrodes and used as biomarkers of the aDBS accurately reflect the true bioelectric signals and avoid erroneous measurements that may adversely affect the therapy provided to the patient. Accordingly, the techniques disclosed herein may provide patients with more effective aDBS therapy than conventional systems.

Fig. 3 is a block diagram of external programmer 104 of fig. 1. Although programmer 104 may be generally described as a handheld device, programmer 104 may be a larger portable device or a more stationary device. Further, in other examples, programmer 104 may be included as part of or include the functionality of an external charging device. As shown in fig. 3, programmer 104 may include processing circuitry 310, memory 311, user interface 302, telemetry circuitry 308, and power supply 320. Memory 311 may store instructions that, when executed by processing circuit 310, cause processing circuit 310 and external programmer 104 to provide functionality attributed throughout this disclosure to external programmer 104. Each of these components or modules may include circuitry configured to perform some or all of the functions described herein. For example, the processing circuitry 310 may include processing circuitry configured to perform the processes discussed with respect to the processing circuitry 310.

Generally, programmer 104 includes any suitable arrangement of hardware, alone or in combination with software and/or firmware, to perform techniques ascribed to programmer 104 and to processing circuitry 310, user interface 302, and telemetry circuitry 308 of programmer 104. In various examples, programmer 104 may include one or more processors, which may include fixed function processing circuitry and/or programmable processing circuitry, as formed by, for example, one or more microprocessors, DSPs, ASICs, FPGAs, or any other equivalent integrated or discrete logic circuitry, as well as any combinations of such components. In various examples, programmer 104 may also include memory 311 (such as RAM, ROM, PROM, EPROM, EEPROM, flash memory, hard disk, CD-ROM) that includes executable instructions for causing the one or more processors to perform the actions attributed to them. Further, although processing circuit 310 and telemetry circuit 308 are described as separate modules, in some examples, processing circuit 310 and telemetry circuit 308 may be functionally integrated with one another. In some examples, the processing circuit 310 and the telemetry circuit 308 correspond to respective hardware units, such as ASICs, DSPs, FPGAs, or other hardware units.

Memory 311 (e.g., a storage device) may store instructions that, when executed by processing circuit 310, cause processing circuit 310 and programmer 104 to provide functionality attributed throughout this disclosure to programmer 104. For example, memory 311 may include instructions that cause processing circuitry 310 to obtain a set of parameters from memory, select a spatial electrode motion pattern, or receive user input and send corresponding commands to IMD 104, or for any other function. Further, memory 311 may include a plurality of programs, wherein each program includes a set of parameters defining a stimulation therapy.

The user interface 302 may include buttons or a keypad, lights, a speaker for voice commands, a display such as a Liquid Crystal Display (LCD), Light Emitting Diodes (LEDs), or Organic Light Emitting Diodes (OLEDs). In some examples, the display may be a touch screen. The user interface 302 may be configured to display any information related to delivery of stimulation therapy, identified patient behavior, sensed patient parameter values, patient behavior criteria, or any other such information. The user interface 302 may also receive user input via the user interface 302. The input may be in the form of, for example, pressing a button on a keypad or selecting an icon from a touch screen.

Telemetry circuitry 308 may support wireless communication between IMD106 and programmer 104 under the control of processing circuitry 310. The telemetry circuit 308 may also be configured to communicate with another computing device via wireless communication techniques or directly with another computing device through a wired connection. In some examples, the telemetry circuitry 308 provides wireless communication via RF or proximal inductive media. In some examples, the telemetry circuit 308 includes an antenna, which may take a variety of forms, such as an internal antenna or an external antenna.

Examples of local wireless communication techniques that may be used to facilitate communication between programmer 104 and IMD106 include RF communication according to the 802.11 or bluetooth specification set or other standard or proprietary telemetry protocols. In this way, other external devices may be able to communicate with programmer 104 without establishing a secure wireless connection. As described herein, telemetry circuitry 308 may be configured to transmit spatial electrode motion patterns or other stimulation parameter values to IMD106 to deliver stimulation therapy.

In some examples, processing circuitry 310 of external programmer 104 defines parameters of the electrical stimulation therapy stored in memory 311 to deliver the adaptive DBS to patient 112. In one example, processor 311 of external programmer 104 issues commands to IMD106 via telemetry circuitry 308 to cause IMD106 to deliver electrical stimulation therapy via electrodes 116, 118 via leads 114.

In accordance with techniques of this disclosure, external programmer 104 may use probability entropy of one or more bioelectrical signals of brain 120 of patient 112 sensed via electrodes 116, 118 of IMD106 to distinguish between electrodes capable of sensing clean bioelectrical signals and electrodes contaminated with artifacts. The techniques of this disclosure recognize that neuronal LFP activity in the brain 120 of the patient 112 is typically a random (e.g., stochastic) process and exhibits high entropy. In contrast, the techniques of this disclosure recognize that patterns or orders (e.g., low entropy) in the sensed bioelectrical signal may indicate artifacts in the sensed signal, such as artifacts that occur due to ECG, motion, or other periodic artifacts. In some examples, external programmer 104 may verify that the bioelectrical signals sensed by one of electrodes 116, 118 of IMD106 are of sufficient quality to be used as biomarkers for controlling the adss therapy. In some examples, the external programmer 104 may use the probability entropy of the bioelectrical signal sensed by one of the recording electrodes 116, 118 as an indicator of ECG or other periodic artifacts present in the recorded LFP signal of the brain 120 of the patient 112.

In one example, IMD106 senses a plurality of bioelectrical signals of brain 120 of patient 112 via electrodes 116, 118. Processing circuitry 310 receives a plurality of bioelectrical signals of brain 120 of patient 112 via telemetry circuitry 308 and from IMD 106. The processing circuitry 310 determines a probability entropy value for the bioelectric signals for each bioelectric signal. Processing circuitry 310 compares each of the respective probability entropy values of the sensed bioelectrical signals to a respective entropy threshold. In some examples, the processing circuitry 310 may determine a probability entropy value for each bioelectric signal and compare the probability entropy to a respective entropy threshold in a manner similar to that described above with respect to the processing circuitry 210 of the IMF 106 of fig. 2.

The processing circuit 310 selects a subset of the electrodes 116, 118 based on the comparison. In some examples, the subset of electrodes 116, 118 includes only those electrodes whose corresponding bioelectric signals are determined by the processing circuitry 310 to indicate no or minimal artifacts. In other words, its corresponding bioelectric signal is determined by the processing circuitry 310 to indicate that the electrode with the artifact is excluded from the subset. Processing circuitry 310 controls IMD106 to deliver electrical stimulation therapy to patient 112 based on the bioelectrical signals sensed via respective electrodes of the subset of electrodes 116, 118 and excludes bioelectrical signals of the plurality of bioelectrical signals sensed via respective electrodes 116, 118 that are not in the subset. In some examples, processing circuitry 310 outputs for display to a clinician or patient a notification of the presence of artifacts in electrodes excluded from the subset.

Fig. 4 is a graphical representation of a sensed bioelectrical signal 400 of a patient. For convenience, FIG. 4 is described with respect to FIG. 1. For example, IMD106 of fig. 1 may sense a bioelectrical signal 400 from brain 120 of patient 112 via one of electrodes 116, 118. In one example, the bioelectric signal 400 may be a sensed LFP of the brain 120 of the patient 112.

In examples where the bioelectric signal 400 is sensed neuronal LFP activity in the brain 120 of the patient 112, the neuronal LFP activity is typically a random (e.g., random) process and exhibits high entropy. In contrast, patterns or orders (e.g., low entropy) occurring in the bioelectric signal 400 may indicate artifacts in the sensed signal, such as artifacts occurring due to ECG, motion, or other periodic artifacts. Thus, the probability entropy of the bioelectric signal 400 can be used to determine whether artifacts are present in the bioelectric signal 400.

For example, fig. 4 depicts threshold limit 404. In some examples, the threshold limit 404 is the first quartile range of the sensed amplitude of the bioelectric signal 400. In some examples, the threshold limit 404 is a second quartile range of the sensed amplitude of the bioelectrical signal 400. The bioelectric signal 400 exhibits a plurality of threshold crossings 402. Each threshold crossing 402 corresponds to an instance in which the amplitude of the bioelectrical signal 400 exceeds a threshold limit 404. Further, the interval length 406 depicts a time interval between a first threshold crossing 402 where the amplitude of the bioelectrical signal 400 exceeds the threshold limit 404 and a second threshold crossing 402 where the amplitude of the bioelectrical signal 400 exceeds the threshold limit 404.

In some examples, IMD106 may calculate a rate over time at which the amplitude of bioelectrical signal 400 exceeds threshold limit 404 by calculating the number of threshold crossings 402 over a period of time (e.g., 1 second, 30 seconds, 60 seconds, etc.). Further, IMD106 compares the rate at which the threshold crosses 402 to the rate threshold. If the rate at which the threshold of the bioelectric signal 400 crosses 402 is less than the rate threshold, the rate at which the threshold crosses 402 may indicate that the bioelectric signal 400 has no or little artifact. In contrast, the rate at which the threshold of the bioelectric signal 400 crosses 402 is greater than the rate threshold, the rate at which the threshold crosses 402 may indicate that the bioelectric signal 400 is contaminated by artifacts.

As another example, IMD106 may calculate interval length 406 by determining a length of time between a first threshold crossing 402 where the amplitude of bioelectrical signal 400 exceeds threshold limit 404 and a second threshold crossing 402 where the amplitude of bioelectrical signal 400 exceeds threshold limit 404. In addition, IMD106 compares the entropy of the plurality of interval lengths 406 to an entropy threshold. If the entropy of the plurality of interval lengths 406 is less than the entropy threshold, the entropy indicates a bioelectric signal contaminated by artifacts. Further, if the entropy of the plurality of interval lengths 406 is greater than the entropy threshold, the entropy indicates a bioelectric signal with no or little artifact.

Fig. 5 is a flow chart illustrating exemplary operations in accordance with the techniques of this disclosure. For convenience, FIG. 5 is described with respect to FIG. 1. In the exemplary operation of fig. 5, IMD106 uses probabilistic entropy of bioelectric signals sensed via electrodes 116, 118 of brain 120 of patient 112 to distinguish between electrodes capable of sensing clean bioelectric signals and electrodes contaminated with artifacts.

In one example, IMD106 senses a plurality of bioelectrical signals of brain 120 of patient 112 via electrodes 116, 118 and sensing circuitry 204 (502). IMD106 determines a probability entropy value for the bioelectrical signals for each bioelectrical signal sensed at the respective electrode 116, 118 (504). In some examples, the probability entropy value of the bioelectric signal is a probability distribution of values of the bioelectric signal over a period of time. In some examples, the probability entropy value of the bioelectric signal is a statistical measure of the randomness of the value of the bioelectric signal over a period of time. Thus, the probability entropy value of the bioelectric signal is a measure of the degree of randomness of the bioelectric signal.

IMD106 compares each of the respective probability entropy values of the sensed bioelectrical signals to a respective entropy threshold (506). IMD106 selects a subset of electrodes 116, 118 based on the comparison (508). In some examples, the subset of electrodes 116, 118 includes only those electrodes whose corresponding bioelectric signals were determined by IMD106 to indicate no or minimal artifacts. In other words, its corresponding bioelectric signal is determined by IMD106 to indicate that electrodes with artifacts are excluded from the subset. IMD106 controls delivery of electrical stimulation therapy to patient 112 based on the bioelectrical signals sensed via respective electrodes of the subset of electrodes 116, 118 and excludes bioelectrical signals of the plurality of bioelectrical signals that are sensed via respective electrodes 116, 118 that are not in the subset (510).

Fig. 6 is a flow chart illustrating exemplary operations in accordance with the techniques of this disclosure. For convenience, FIG. 6 is described with respect to FIG. 1. In the exemplary operation of fig. 6, IMD106 determines probability entropy values for the bioelectrical signals sensed from patient 112 by analyzing statistical measures of randomness of spectral power over multiple frequency bands of the bioelectrical signals.

For example, IMD106 splits the bioelectric signals into multiple frequency bands. IMD106 determines a statistical measure of randomness of the spectral power over multiple frequency bands (602). IMD106 compares the statistical measure of randomness of the spectral power over the plurality of frequency bands to the respective entropy thresholds (604). As one example, if a statistical measure of randomness of the power of the first frequency band exceeds an entropy threshold of the first frequency band, the statistical measure of randomness indicates a bioelectric signal with no or little artifact. As another example, if the statistical measure of randomness of the power of the second frequency band is less than the entropy threshold of the second frequency band, the statistical measure of randomness indicates a bioelectric signal that is not contaminated by artifacts. In some examples, if one statistical measure of randomness of the power of one frequency band indicates that the bioelectric signal is contaminated by artifacts, IMD106 determines that the bioelectric signal is contaminated by artifacts. In another example, IMD106 determines that the bioelectrical signal is contaminated by the artifact only if each statistical measure of randomness of the power of each frequency band within the LFP recording indicates that the bioelectrical signal is contaminated by the artifact.

IMD106 selects a subset of electrodes 116, 118 based on the comparison (606). For example, IMD106 selects each of electrodes 116, 118 that satisfies the following condition: for each of the electrodes, a statistical measure of randomness of spectral power over a plurality of frequency bands of the respective bioelectrical signal sensed via the electrode is greater than a respective entropy threshold.

Fig. 7 is a flow chart illustrating exemplary operations in accordance with the techniques of this disclosure. For convenience, FIG. 7 is described with respect to FIG. 1. In the exemplary operation of fig. 7, IMD106 determines a probability entropy value for the bioelectrical signal sensed from patient 112 by analyzing a rate over time at which the amplitude of the bioelectrical signal exceeds a threshold limit.

In one example, IMD106 determines a rate over time at which the amplitude of the bioelectrical signal exceeds a threshold limit (702). In some examples, the threshold limit is a first quartile range of the sensed amplitude of the bioelectric signal. IMD106 compares the determined rate to a rate threshold of the bioelectrical signal (704). As one example, IMD106 determines a rate over time at which the amplitude of the first bioelectrical signal exceeds a threshold limit and a rate over time at which the amplitude of the second bioelectrical signal exceeds a threshold limit. If the rate of the first bioelectric signal is less than the rate threshold, then IMD106 determines that the rate of the first bioelectric signal is indicative of a no artifact or a nearly artifact free bioelectric signal. Further, if the rate of the second bioelectric signal is greater than the rate threshold, the IMD106 determines that the rate of the second bioelectric signal is indicative of a bioelectric signal contaminated by the artifact.

IMD106 selects a subset of electrodes 116, 118 based on the comparison (706). For example, IMD106 selects each of electrodes 116, 118 that satisfies the following condition: for each of the electrodes, a rate at which an amplitude of the respective bioelectrical signal sensed via the electrode exceeds a threshold limit is less than the respective rate threshold.

Fig. 8 is a flow chart illustrating exemplary operations in accordance with the techniques of this disclosure. For convenience, FIG. 8 is described with respect to FIG. 1. In the exemplary operation of fig. 8, IMD106 determines a probability entropy value for the bioelectric signal sensed from patient 112 by analyzing entropy for the length of consecutive intervals between instances in which the amplitude of the bioelectric signal exceeds a threshold limit.

In one example, IMD106 determines an entropy of a plurality of time intervals between a first instance in which the amplitude of the bioelectrical signal exceeds a threshold limit and a second instance in which the amplitude of the bioelectrical signal exceeds the threshold limit (802). In some examples, IMD10 determines shannon entropy for the time interval. In some examples, the threshold limit is a first quartile range of the sensed amplitude of the bioelectric signal.

IMD106 compares the entropy of the time interval to an entropy threshold of the bioelectrical signal (804). For example, IMD106 identifies a time interval between a first instance in which the amplitude of the bioelectrical signal exceeds a threshold limit and a second instance in which the amplitude of the bioelectrical signal exceeds the threshold limit. Further, IMD106 determines the entropy of the time interval between successive instances in which the amplitude of the bioelectrical signal exceeds the threshold limit. If the entropy of the interval is less than the entropy threshold, then IMD106 determines that the entropy of the interval is indicative of a bioelectrical signal contaminated by the artifact. Further, if the entropy of the interval is greater than the entropy threshold, IMD106 determines that the interval indicates a bioelectric signal with no or little artifact.

IMD106 selects a subset of electrodes 116, 118 based on the comparison (806). For example, IMD106 determines, for each of electrodes 116, 118, an entropy of a time interval between successive instances in which an amplitude of a respective bioelectrical signal sensed via the electrode exceeds a signal threshold limit. IMD106 selects each of electrodes 116, 118 that satisfies the following condition: for each of the electrodes, the entropy of successive intervals is greater than a respective entropy threshold.

The following examples may illustrate one or more aspects of the present disclosure.

Embodiment 1. a method comprising: sensing a plurality of bioelectrical signals of a brain of a patient via a plurality of electrodes; determining, by the processing circuitry and for each bioelectrical signal of the plurality of bioelectrical signals sensed at a respective electrode of the plurality of electrodes, a probability entropy value for the bioelectrical signal; comparing, by the processing circuit, each of the respective probability entropy values of the bioelectric signals to a respective entropy threshold; and selecting, by the processing circuitry and based on the comparison, a subset of the electrodes of the plurality of electrodes; and controlling, by the processing circuitry and based on and excluding ones of the plurality of bioelectrical signals sensed via respective electrodes of the subset of electrodes, delivery of the electrical stimulation therapy to the patient.

Embodiment 2. the method of embodiment 1, wherein determining the probability entropy value for each bioelectric signal comprises determining a probability distribution of entropy of the bioelectric signal over a period of time for each bioelectric signal, wherein signal components of the bioelectric signal having periodic behavior are indicative of decreasing entropy, and wherein signal components of the bioelectric signal not having periodic behavior are indicative of increasing entropy.

Embodiment 3. the method of any of embodiments 1 to 2, wherein determining a probability entropy value for each bioelectric signal comprises: determining that one or more signal components of a first bioelectric signal of the plurality of bioelectric signals exhibit periodic behavior; determining a first probability entropy value for the first bioelectric signal in response to determining that one or more signal components of the first bioelectric signal exhibit periodic behavior; determining that one or more signal components of a second bioelectric signal of the plurality of bioelectric signals exhibit aperiodic behavior; and in response to determining that one or more signal components of the second bioelectric signal exhibit aperiodicity, determining a second probability entropy value for the second bioelectric signal, wherein the first probability entropy value is indicative of entropy in the first bioelectric signal and the second probability entropy value is indicative of entropy in the second bioelectric signal, and wherein the first probability entropy value and the second probability entropy value indicate that the first bioelectric signal exhibits less entropy than the second bioelectric signal.

Embodiment 4. the method of any of embodiments 1 to 3, wherein determining the probability entropy value for each bioelectric signal comprises determining a statistical measure of randomness over a period of time for each bioelectric signal.

Embodiment 5. the method of embodiment 4, wherein determining, for each bioelectrical signal, a statistical measure of randomness comprises determining a statistical measure of randomness of spectral power over a plurality of frequency bands of the bioelectrical signal, wherein comparing each of the respective probability entropy values of the bioelectrical signal to a respective entropy threshold comprises comparing each of the statistical measures of randomness of spectral power over the plurality of frequency bands of the bioelectrical signal to an entropy threshold, and wherein selecting, based on the comparison, a subset of the electrodes of the plurality of electrodes comprises selecting each of the plurality of electrodes that satisfies the following condition: for each of the electrodes, a statistical measure of randomness of spectral power over a plurality of frequency bands of the respective bioelectrical signal sensed via the electrode is greater than a respective entropy threshold.

Embodiment 6. the method of any of embodiments 1 to 5, wherein comparing each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold comprises: determining a rate at which the amplitude of each bioelectric signal exceeds a threshold limit; and comparing the rate at which the amplitude of the bioelectrical signal exceeds a threshold limit with a respective rate threshold, and wherein selecting a subset of the electrodes of the plurality of electrodes based on the comparison comprises selecting each electrode of the plurality of electrodes that satisfies the following condition: for each of the electrodes, a rate at which an amplitude of the respective bioelectrical signal sensed via the electrode exceeds a threshold limit is less than the respective rate threshold.

Embodiment 7. the method of embodiment 6, wherein each threshold limit is a quartile distance of the sensed amplitude of the respective bioelectric signal.

Embodiment 8 the method of any one of embodiments 1 to 2, wherein comparing each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold comprises: determining an entropy of a time interval between instances where the amplitude of the bioelectrical signal exceeds a threshold limit; and comparing the entropy of the time interval to respective entropy thresholds, and wherein selecting a subset of the electrodes of the plurality of electrodes based on the comparison comprises selecting each electrode of the plurality of electrodes that satisfies the following condition: for each of the electrodes, the entropy of the time interval is greater than the respective entropy threshold.

Embodiment 9. the method of embodiment 8, wherein each threshold limit is a first quartile of the sensed amplitude of the respective bioelectric signal.

Embodiment 10 the method of any one of embodiments 8 to 9, wherein determining the time interval comprises determining a shannon entropy of the time interval.

Embodiment 11 the method of any one of embodiments 1 to 10, wherein comparing each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold comprises: determining a statistical measure of randomness of the spectral power over a plurality of frequency bands for each bioelectric signal; comparing the statistical measure of randomness of spectral power over a plurality of frequency bands of the bioelectric signal to a first entropy threshold; determining a rate at which the amplitude of the bioelectrical signal exceeds an amplitude threshold limit; comparing a rate at which the amplitude of the bioelectrical signal exceeds an amplitude threshold limit to a rate threshold; determining an entropy of a time interval between instances where the amplitude of the bioelectrical signal exceeds a threshold limit; comparing the entropy of the time interval to a second entropy threshold, and wherein selecting the subset of electrodes of the plurality of electrodes based on the comparison comprises selecting each electrode of the subset of electrodes of the plurality of electrodes based on a statistical measure of randomness of spectral power over a plurality of frequency bands of the respective bioelectrical signals sensed via the electrodes, a rate at which amplitudes of the respective bioelectrical signals sensed via the electrodes exceed an amplitude threshold limit, and an entropy of the time interval between instances at which the amplitudes of the bioelectrical signals sensed via the electrodes exceed the threshold limit.

Embodiment 12 the method of embodiment 11, wherein comparing each of the respective probability entropy values of the bioelectrical signal to a respective entropy threshold, and selecting the subset of electrodes of the plurality of electrodes based on the comparison further comprises: processing, by a machine learning system executing on processing circuitry, a statistical measure of randomness of spectral power over a plurality of frequency bands of respective bioelectrical signals sensed via electrodes, a rate at which amplitudes of the respective bioelectrical signals sensed via the electrodes exceed an amplitude threshold limit, and an entropy of a time interval between instances at which the amplitudes of the bioelectrical signals sensed via the electrodes exceed the threshold limit to identify one or more electrodes of the plurality of electrodes, and selecting, by the machine learning system, the identified one or more electrodes as a subset of the electrodes of the plurality of electrodes.

Embodiment 13. the method of any of embodiments 1 to 12, wherein comparing each of the respective probability entropy values of the bioelectrical signal to a respective entropy threshold, and selecting the subset of the electrodes of the plurality of electrodes based on the comparison comprises: processing, by a machine learning system executing on the processing circuit, each of the respective probability entropy values of the bioelectrical signals to identify one or more electrodes of the plurality of electrodes, and selecting, by the machine learning system, the identified one or more electrodes as a subset of the electrodes of the plurality of electrodes.

Embodiment 14 the method of any one of embodiments 1 to 13, wherein comparing each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold comprises: determining a normalization of each bioelectrical signal; and comparing the normalized probability entropy of the bioelectrical signal with a respective entropy threshold.

Embodiment 15. the method of any of embodiments 1 to 14, further comprising: for each electrode not in the subset of electrodes, outputting, by the processing circuitry, an indication that an artifact is present in the respective bioelectric signal sensed via the electrode.

Embodiment 16 an implantable medical device, comprising: a plurality of electrodes; sensing circuitry configured to sense a plurality of bioelectrical signals of a brain of a patient via a plurality of electrodes; and processing circuitry configured to: determining, for each bioelectric signal of a plurality of bioelectric signals sensed at a respective electrode of a plurality of electrodes, a probability entropy value for the bioelectric signal; comparing each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold; selecting a subset of electrodes of the plurality of electrodes based on the comparison; and controlling delivery of the electrical stimulation therapy to the patient based on and excluding ones of the plurality of bioelectrical signals sensed via respective electrodes of the subset of electrodes.

Embodiment 17 the implantable medical device of embodiment 16, wherein to determine the probabilistic entropy value of the bioelectric signal, the processing circuitry is further configured to determine a probability distribution of entropy of the bioelectric signal over a period of time, wherein signal components of the bioelectric signal having periodic behavior are indicative of decreased entropy, and wherein signal components of the bioelectric signal not having periodic behavior are indicative of increased entropy.

Embodiment 18 the implantable medical device of any one of embodiments 16-17, wherein to determine the probability entropy value for each bioelectric signal, the processing circuitry is further configured to: determining that one or more signal components of a first bioelectric signal of the plurality of bioelectric signals exhibit periodic behavior; determining a first probability entropy value for the first bioelectric signal in response to determining that one or more signal components of the first bioelectric signal exhibit periodic behavior; determining that one or more signal components of a second bioelectric signal of the plurality of bioelectric signals exhibit aperiodic behavior; and in response to determining that one or more signal components of the second bioelectric signal exhibit aperiodic behavior, determining a second probability entropy value for the second bioelectric signal, wherein the first probability entropy value is indicative of entropy in the first bioelectric signal and the second probability entropy value is indicative of entropy in the second bioelectric signal, and wherein the first probability entropy value and the second probability entropy value indicate that the first bioelectric signal exhibits less entropy than the second bioelectric signal.

Embodiment 19 the implantable medical device of any one of embodiments 16-18, wherein to determine the probability entropy value of the bioelectrical signal, the processing circuitry is further configured to determine a statistical measure of randomness over a period of time.

Embodiment 20 the implantable medical device of embodiment 19, wherein to determine the probability entropy values of the bioelectric signal, the processing circuitry is further configured to determine statistical measures of randomness of spectral power over a plurality of frequency bands of the bioelectric signal, wherein to compare each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold, the processing circuitry is further configured to compare each statistical measure of randomness of spectral power over a plurality of frequency bands of the bioelectric signal to an entropy threshold, and wherein to select the subset of electrodes of the plurality of electrodes based on the comparison, the processing circuitry is further configured to select each electrode of the plurality of electrodes that satisfies the following condition: for each of the electrodes, a statistical measure of randomness of spectral power over a plurality of frequency bands of the respective bioelectrical signal sensed via the electrode is greater than a respective entropy threshold.

Embodiment 21 the implantable medical device of any one of embodiments 16-20, wherein to compare each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold, the processing circuitry is further configured to: determining a rate at which the amplitude of each bioelectric signal exceeds a threshold limit; and comparing the rate at which the amplitude of the bioelectric signal exceeds a threshold limit with a corresponding rate threshold; and wherein to select a subset of the electrodes of the plurality of electrodes based on the comparison, the processing circuitry is further configured to select each electrode of the plurality of electrodes that satisfies the following condition: for each of the electrodes, a rate at which an amplitude of the respective bioelectrical signal sensed via the electrode exceeds a threshold limit is less than the respective rate threshold.

Embodiment 22 the implantable medical device of any one of embodiments 16-21, wherein to compare each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold, the processing circuitry is further configured to: determining an entropy of a time interval between instances where the amplitude of the bioelectrical signal exceeds a threshold limit; and comparing the entropy of the time interval to respective entropy thresholds, and wherein to select a subset of the electrodes of the plurality of electrodes based on the comparison, the processing circuitry is further configured to select each electrode of the plurality of electrodes that satisfies the following condition: for each of the electrodes, the entropy of the time interval is greater than the respective entropy threshold.

Embodiment 23. a system, comprising: an implantable medical device, comprising: a plurality of electrodes; sensing circuitry configured to sense a plurality of bioelectrical signals of a brain of a patient via a plurality of electrodes; and processing circuitry configured to: determining a probability entropy value for the bioelectric signals for each bioelectric signal of the plurality of bioelectric signals sensed at a respective electrode of the plurality of electrodes; comparing each of the respective probability entropy values of the bioelectric signal to a respective entropy threshold; selecting a subset of electrodes of the plurality of electrodes based on the comparison; and controlling delivery of electrical stimulation therapy to the patient based on and excluding ones of the plurality of bioelectrical signals sensed via respective electrodes of the subset of electrodes.

Embodiment 24 the system of embodiment 23, wherein the implantable medical device comprises a processing circuit.

Embodiment 25. the system of any of embodiments 23 to 24, further comprising an external device, wherein the external device comprises the processing circuitry.

The techniques described in this disclosure may be implemented, at least in part, in hardware, software, firmware, or any combination thereof. For example, various aspects of the described techniques may be implemented within one or more processors, including one or more microprocessors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs), or any other equivalent integrated or discrete logic circuitry, as well as any combinations of such components. The term "processor" or "processing circuitry" may generally refer to any of the foregoing logic circuitry, alone or in combination with other logic circuitry, or any other equivalent circuitry. The control unit, including hardware, may also perform one or more of the techniques of this disclosure.

Such hardware, software, and firmware may be implemented within the same device or within separate devices to support the various operations and functions described in this disclosure. Further, any of the units, modules or components may be implemented together or separately as discrete but interoperable logic devices. The description of different features as modules or units is intended to highlight different functional aspects and does not necessarily imply that such modules or units must be realized by separate hardware or software components. Rather, functionality associated with one or more modules or units may be performed by separate hardware or software components, or integrated within common or separate hardware or software components.

The techniques described in this disclosure may also be embedded or encoded in a computer-readable medium (such as a computer-readable storage medium) containing instructions. Instructions embedded or encoded in a computer-readable storage medium may cause a programmable processor or other processor, for example, to perform the method when executing the instructions. The computer-readable storage medium may include Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read Only Memory (PROM), Erasable Programmable Read Only Memory (EPROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory, a hard disk, a CD-ROM, a floppy disk, a magnetic tape cartridge, magnetic media, optical media, or other computer-readable medium.