阅读说明：本技术 用于监测的集成可穿戴超声相控阵列 (Integrated wearable ultrasound phased array for monitoring ) 是由 徐升 林沐阳 张倬瑞 胡鸿杰 王冲和 齐白岩 于 2020-02-28 设计创作，主要内容包括：提供了将控制电子设备与无线板载模块集成在一起的系统和方法,使得共形超声装置是功能齐全且独立的系统。此类系统采用集成控制电子设备、深层组织监测、无线通信和智能机器学习算法来分析数据。特别地,提供了执行所述功能的可伸缩超声贴片。解码的运动信号可能对血压估计、慢性阻塞性肺病(COPD)诊断、心脏功能评估和许多其他医疗监测方面有影响。(Systems and methods are provided that integrate control electronics with a wireless on-board module such that the conformal ultrasound device is a fully functional and self-contained system. Such systems employ integrated control electronics, deep tissue monitoring, wireless communication, and intelligent machine learning algorithms to analyze the data. In particular, a retractable ultrasound patch is provided that performs the described functions. The decoded motion signals may have an impact on blood pressure estimation, Chronic Obstructive Pulmonary Disease (COPD) diagnosis, cardiac function assessment and many other medical monitoring aspects.)

1. A system for physiological parameter monitoring, comprising:

a. a conformal ultrasound transducer array coupled to the flexible substrate;

b. analog front end circuitry coupled to the flexible substrate and further coupled to the conformal ultrasonic transducer array, the analog front end circuitry configured to generate ultrasonic waves and receive reflected ultrasonic waves;

c. digital circuitry coupled to the flexible substrate and further coupled to the analog front end circuitry, the digital circuitry configured to at least:

i. controlling the analog front-end circuit at least in ultrasonic wave generation;

transmitting an indication of the received reflected ultrasound waves to an external computing environment.

2. The system of claim 1, further comprising the external computing environment.

3. The system of claim 1, wherein the external computing environment is configured to generate and display an indication of the function of the monitored organ.

4. The system of claim 1, wherein the external computing environment is configured to measure a shift, the shift being a shift in the time domain, a shift in a detected peak of the received reflected acoustic wave, a shift due to movement of an organ or tissue, wherein the displayed indication of the monitored physiological parameter is based on the measured shift.

5. The system of claim 4, wherein the identification of the offset is based at least in part on a step of machine learning.

6. The system of claim 5, wherein the displayed indication is based on a step of machine learning that associates the offset with the monitored physiological parameter.

7. The system of claim 1, wherein the analog front end is further configured to direct or orient the generated ultrasound waves toward an organ, tissue, or location of interest, the directing or orienting being by beamforming.

8. The system of claim 7, wherein the directing comprises dynamically adjusting a time delay profile of individual transducer activations in the transducer array.

9. The system of claim 1, wherein the flexible substrate is made of polyimide.

10. The system of claim 1, wherein the transducer array comprises a piezoelectric array.

11. The system of claim 1, wherein the monitored physiological parameter is central blood pressure or COPD.

12. A method for monitoring a physiological parameter, comprising:

a. determining a location of interest, the location being associated with a physiological parameter to be monitored;

b. transmitting ultrasound waves to the location of interest;

c. receiving reflected ultrasound waves from the location of interest;

d. transmitting an indication of the received reflected ultrasonic waves to an external computing environment;

e. receiving the received reflected ultrasonic waves at the external computing environment;

f. detecting a time domain offset of the received reflected ultrasound waves;

g. determining an indication of the monitored physiological parameter based at least in part on the offset; and

h. displaying an indication of the monitored physiological parameter;

i. wherein at least transmitting and receiving reflected ultrasound waves and transmitting the indication are performed by components within the integrated wearable device.

13. The method of claim 11, wherein the monitored physiological parameter is central blood pressure.

14. The method of claim 11, wherein said transmitting ultrasound waves to said location of interest comprises the step of directing said ultrasound waves towards said location of interest.

15. The method of claim 14, wherein the steering comprises dynamically adjusting a time delay profile of individual transducer activations in the transducer array.

16. The method of claim 11, wherein transmitting and receiving ultrasound is performed at least in part by a piezoelectric array.

17. The method of claim 11, wherein said detecting a shift in the received reflected ultrasound waves, i.e., a shift in the peak in the time domain, comprises the step of identifying said shift using machine learning.

18. The method of claim 11, wherein said determining an indication of the monitored physiological parameter based, at least in part, on the offset comprises the step of using machine learning to associate the offset with the physiological parameter.

19. The method of claim 16, wherein the machine learning learns on a training set of ultrasound data.

Background

It is known to measure blood pressure in various ways. The standard method is to use a blood pressure cuff. Alternative and more advanced methods have also been developed.

For example, PCT/US2018/013116 entitled "scalable ultrasound transducer apparatus" describes a skin-integrated conformal ultrasound apparatus capable of non-invasively acquiring Central Blood Pressure (CBP). The system requires connecting the ultrasound patch to a background data acquisition system. Although useful, it has the disadvantage of requiring such data coupling.

This background is provided to introduce a brief summary of the disclosure and detailed description that follows. This background is not intended to be an aid in determining the scope of the claimed subject matter, nor is it to be construed as limiting the claimed subject matter to implementations that solve any or all of the disadvantages or problems described above.

Disclosure of Invention

Systems and methods in accordance with the present principles address the above needs in a variety of ways.

In particular, there is a need to integrate control electronics with a wireless on-board module so that the conformal ultrasound device becomes a fully functional, self-contained system. This provides an important step in the conversion of the system from table top to bedside. Such systems may employ integrated control electronics, deep tissue monitoring, wireless communication, and intelligent machine learning algorithms to analyze the data.

In one aspect, methods, apparatuses, and systems related to a fully integrated smart wearable ultrasound system are disclosed. Such systems and methods allow human body biological interface motion monitoring through a retractable ultrasound patch. Decoded motion signals may have an impact on blood pressure estimation, Chronic Obstructive Pulmonary Disease (COPD) diagnosis, cardiac function assessment, and many other medical monitoring aspects.

In one aspect, the present invention relates to a physiological parameter monitoring system comprising: a conformal ultrasound transducer array coupled to the flexible substrate; analog front end circuitry coupled to the flexible substrate and further coupled to the conformal ultrasonic transducer array, the analog front end circuitry configured to generate ultrasonic waves and receive reflected ultrasonic waves; a digital circuit coupled to the flexible substrate and further coupled to the analog front end circuit, the digital circuit configured to at least: controlling the analog front-end circuit to at least generate ultrasonic waves; transmitting an indication of the received reflected ultrasonic waves to an external computing environment.

Embodiments of the invention may include one or more of the following. The system may further include the external computing environment, which may be configured to generate and display an indication of the function of the monitored organ. The external computing environment may also be configured to measure a shift, the shift in the time domain, the shift in the detected peak of the received reflected acoustic wave, the shift due to movement of an organ or tissue, wherein the displayed indication of the monitored physiological parameter is based on the measured shift. The identification of the offset is based at least in part on the step of machine learning. The displayed indication is based on a step of machine learning that associates the offset with the monitored physiological parameter. The analog front end is further configured to direct or steer the generated ultrasound waves to an organ, tissue, or location of interest, the steering or steering being by beamforming. The steering includes dynamically adjusting a time delay profile of individual transducer activations in the transducer array, which includes a piezoelectric array. The flexible substrate is made of polyimide. The monitored physiological parameter is central blood pressure or COPD.

In another aspect, the invention relates to a method for monitoring a physiological parameter, comprising: determining a location of interest, the location being related to a physiological parameter to be monitored; transmitting ultrasound waves to the location of interest; receiving ultrasound waves reflected from the location of interest; transmitting an indication of the received reflected ultrasonic waves to an external computing environment; receiving the received reflected ultrasonic waves at the external computing environment; detecting a time domain offset of the received reflected ultrasound waves; determining an indication of the monitored physiological parameter based at least in part on an offset; and displaying an indication of the monitored physiological parameter; wherein at least transmitting and receiving reflected ultrasound waves and transmitting the indication are performed by components within the integrated wearable device.

Embodiments of the invention may include one or more of the following. The monitored physiological parameter is central blood pressure. Transmitting ultrasound waves to the location of interest includes the step of steering the ultrasound waves toward the location of interest, wherein the steering includes dynamically adjusting a time delay profile of individual transducer activations in the transducer array. Transmitting and receiving ultrasound waves is performed, at least in part, by a piezoelectric array. Detecting a shift in the received reflected ultrasound waves, i.e. the shift in the peak value in the time domain, comprising the step of identifying the shift using machine learning. Determining the indication of the monitored physiological parameter is based at least in part on the offset and includes the step of using machine learning to associate the offset with the physiological parameter. The machine learning is learned on a training set of ultrasound data.

In certain embodiments, advantages of the invention may include one or more of the following. Biomedical imaging as claimed herein is ultrasonically visible imaging including, but not limited to, blood vessel walls, septa, heart valves, and the like. In one aspect, the new ultrasound imaging system overcomes the challenge of locating the uncertain position of the transducer using unsupervised machine learning algorithms, as compared to existing ultrasound imaging probes. In addition, the technique may also perform real-time artificial intelligence (Al) analysis to extract hemodynamic factors, such as blood pressure, blood flow, and cardiac pressure signals, from the ultrasound images. Other advantages will be understood from the following description, including the drawings and claims.

This summary is provided to introduce a selection of concepts in a simplified form. These concepts are further described in the detailed description section. Elements or steps other than those described in this summary are possible and do not require any elements or steps. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter. The claimed subject matter is not limited to implementations that solve any or all disadvantages noted in any part of this disclosure.

Drawings

Fig. 1 shows a schematic diagram of an embodiment in accordance with the present principles.

Figure 2A shows a more detailed schematic diagram of an embodiment in accordance with the present principles.

Fig. 2B illustrates a more detailed implementation of an analog front end in accordance with the present principles.

Fig. 3 shows a more detailed embodiment of an exemplary transducer unit in accordance with the present principles.

Fig. 4 shows an exemplary hardware design (circuit schematic) for a wireless ultrasound front end in accordance with the present principles.

Fig. 5 illustrates the timing control logic of the MCU for pulse generation, rf signal digitization, and data transmission within one pulse repetition interval.

Fig. 6A illustrates a GUI diagram of software in the workflow of an automated signal processing algorithm, taking vasodilation monitoring as an example.

Fig. 6B shows the steps of automatic channel selection and automatic motion tracking.

Fig. 6C shows an exemplary software design for autonomous artery identification and wall tracking.

Fig. 7 shows an example of a peak shift.

FIG. 8A illustrates the use of an unsupervised machine learning algorithm to find transducer positions to improve the quality of the reconstructed image.

Fig. 8B shows a proposed algorithm for ultrasound image quality enhancement.

Fig. 8C schematically shows enhancement of an image.

Fig. 9 illustrates a deep learning architecture (9A) and a two-way domain adaptation method (9B) for ultrasound image interpretation.

Fig. 10A and 10B illustrate the use of a conformal ultrasound patch on a user. Figure 10B also shows the central blood vessel of the human neck.

Fig. 11 shows an exemplary embodiment of a conformal ultrasound transducer array, which is indicative of conforming to a curved body surface.

Fig. 12-15 illustrate an exemplary embodiment of a system and method in accordance with the present principles, particularly a dense array device arranged for imaging and doppler ultrasound.

Fig. 12 illustrates a core technique for receive beamforming.

Fig. 13A and 13B illustrate an application of the technique according to the present principles employed in non-destructive testing.

Fig. 14 illustrates an application of the technique in accordance with the present principles employed in B-mode ultrasound.

Fig. 15 shows a core technique for transmit beamforming.

Figures 16A and 16B illustrate an application of the technique in accordance with the present principles employed in tissue doppler imaging.

Fig. 17A and 17B illustrate the application of techniques in accordance with the present principles in blood flow monitoring.

Like reference numerals refer to like elements throughout. Elements are not drawn to scale unless otherwise indicated.

Detailed Description

Arrangements in accordance with the present principles include materials, devices, systems, and methods related to fully integrated smart wearable ultrasound systems. Depending on the implementation, the following functional modules may be employed.

Referring to fig. 1, a wearable device 100 may include an ultrasound transducer array 102 coupled to an ultrasound Analog Front End (AFE)104 and digital circuitry 106 for control and communication. The wearable device 100 may be coupled to a receiver 200 that includes an analysis system that includes a communication circuit 108 for receiving signals from the digital circuit 106. The receiver 200 also includes a computing environment 112 running interactive software that can communicate with various back-end devices, such as smartphones, to allow visualization of human biological interface motion waveforms. The machine learning algorithm module 114 may also be used for various functions including automatic transducer channel selection and interface motion waveform decoding from ultrasonic RF signals.

The ultrasound transducer array 102 may be a conformal array that transmits ultrasound and receives reflected acoustic signals. The ultrasound analog front end 104 may be used for ultrasound generation, echo signal reception, and amplification. Other components of the AFE include a high voltage pulse generator, a transmit/receive (T/R) switch, a multiplexer, and a Radio Frequency (RF) amplifier.

Digital circuitry 106 may be used for system control, signal digitization, on-board transmission, and high-speed wireless transmission, among other functions that may be desired. Such digital circuitry 106 typically includes a microcontroller unit (MCU) with a built-in analog-to-digital converter (ADC) and Wi-Fi modules.

Various aspects of these modules, and their use in non-invasive measurement of central blood pressure and other applications, will now be described in more detail.

Fig. 2A illustrates the general principle of biological interface motion monitoring, and fig. 2A illustrates a device for tracking the motion of a vessel wall. The ultrasound transducer element 102 above the target biological interface a (103) generates ultrasound 105 and receives reflected signals therefrom. It can be seen that the acoustic waves transmitted by the transducer elements can be directed and aimed at a specific element, for example a beating artery 107.

As these interfaces move, the reflection peaks shift in the time domain corresponding to their motion. All signals are amplified by the AFE104, digitized by ADCs in MCUs within the digital circuitry 106, and wirelessly transmitted to a smartphone or other analysis system 200, which may run software 114. The machine learning algorithm included in the software 114 may be for identifying the reflected signal of the target interface and continuously capturing its motion trajectory. The algorithm may be located on the smartphone or on a connected computing environment (e.g., a cloud server), for example. The algorithm may employ machine learning to identify offsets caused by motion of the location of interest, and may further use machine learning to correlate the offsets with parameters desired to be monitored, such as physiological parameters desired to be determined for diagnostic and other purposes.

In more detail, in a first step, and referring to fig. 2B and 2C, the analog front end circuit 104 coupled to the transducer array 102 includes a multiplexer 136, a high voltage boost pulse generator 134, a Radio Frequency (RF) amplifier 142, a transmit/receive (T/R) switch 138, and an analog-to-digital converter. Multiple frequency channels allow beam steering and the same beam emerges from the boost pulse generator 134 which is controlled by the digital circuitry 106 to generate ultrasound. The echo signals are amplified and collected using a T/R switch 138 and a multiplexer divider 136 and amplifier 142 that form part of the high speed analog to digital converter. The inset shows the signal flow.

Second, the digitized signal is processed by a Field Programmable Gate Array (FPGA) or MCU. The raw ultrasound data may be decoded into a blood pressure waveform. Finally, the decoded waveform may be wirelessly transmitted via bluetooth or Wi-Fi and displayed on a display. A rechargeable micro-battery may provide power to the entire system.

The ultrasonic transmitter is constituted by a voltage boost circuit which converts a low voltage Control Signal (CS) into high voltage pulses. The T/R switch is used for cutting off the over-range voltage and protecting the receiving circuit. The multiplexer is used for channel selection. The RF amplifier amplifies the received Echo Signals (ES) for subsequent ADC sampling. All components can be manufactured on a Flexible Printed Circuit Board (FPCB).

Fig. 2C illustrates another embodiment of a wireless ultrasound front-end circuit with similar components in a similar arrangement.

It can be seen that the hardware interfacing with the soft ultrasound probe can perform transducer selection, transducer activation, echo signal reception, and wireless data transmission. In one embodiment, a High Voltage (HV) switch 147 controlled by a Microcontroller (MCU)149 may select an appropriate number of transducers as active pixels. Once an active pixel is selected, the pulse generator 134 may transmit an electrical pulse to the pixel to generate an ultrasound wave. After the ultrasonic wave is generated, the echo signal can be received. The received signal may be amplified by an RF amplifier 142 through a transmit/receive (T/R) switch 138 and an analog filter 141. Finally, the amplified signal may be received by an analog-to-digital converter (ADC)143, which may also be an MCU. Once the signal is received and digitized, Wi-Fi module 151 can wirelessly transmit the signal to end device (e.g., PC or smartphone) 112.

Details of an exemplary conformal ultrasound transducer array are shown in fig. 3, which illustrates a schematic diagram of the conformal ultrasound transducer array and the structure of the individual transducer elements (inset). In this exemplary embodiment, an "island bridge" structure is used to provide sufficient flexibility to the device to provide a suitable fit to the skin.

Rigid features 116 are integrated with the islands and undulating serpentine metal interconnects 118 act as bridges. The bridge may bend and twist to absorb externally applied strain. Thus, the entire structure is locally rigid in the islands, but can be globally stretched by adjusting the spacing between the rigid islands during bending, stretching, and twisting. The result is a natural interface that can accommodate the geometry and motion of the skin surface with minimal mechanical constraints, thereby creating a strong, non-irritating device/skin contact that bridges the gap between conventional rigid planar high performance electronics and soft curve dynamic living beings. In one embodiment, the ultrasonic transducers as the rigid components 116 are provided on a substrate 120 having vias 122 for interconnection.

As seen in the inset, the exemplary element 116 may employ a 1-3 piezo-electric composite ultrasonic array assembly 124, also referred to as a piezo post, covered by Cu/Zn electrodes 126, the Cu/Zn electrodes 126 being covered on top and bottom surfaces by copper electrodes 128 and having a polyimide covering 132. It should be noted, however, that the active ultrasonic material used herein is not limited to 1-3 composite materials, but any rigid piezoelectric material may be employed. The polyamide layer may provide a substrate as well as a cover.

Fig. 4 illustrates the operating logic of digital circuit 106. As described above, the digital circuitry may include the MCU149, the integrated ADC (e.g., element 143), and the Wi-Fi module 151. Referring now to the figures, for ultrasound transmission, a trigger signal 153 is used to generate an ultrasound pulse in a triggering step 144. After the trigger signal 153, the RF signal 155 of the ultrasonic echo is received by the transducer. In step 146, the ADC is simultaneously activated to digitally sample the received ultrasound echoes. To achieve a sufficient sampling frequency, the embedded ADC may in one implementation operate in an interleaved manner. The designed sampling rate may be proportional to the number of embedded ADCs and the sampling rate of 1. A typical composite sampling rate is 20 MHz. The ADC may operate within a predefined time-gated range and store all data into the MCU's built-in memory. The data may then be wirelessly transmitted to the terminal device via TCP/IP protocol in step 148. Direct Memory Access (DMA) techniques may be employed to guarantee data access speed. The digital circuit may be fabricated on the FPCB platform and integrated into the AFE circuit.

Referring to fig. 5, software 152 may be employed on the terminal device 112, for example on a computing environment such as a smartphone, laptop, tablet, desktop, etc., to receive wirelessly transmitted data from the wearable apparatus 100, to process the data, and to visualize detected biological interface motion (e.g., motion of an artery wall). For example, on a Graphical User Interface (GUI)154, a user may connect the backend terminal 112 to the wearable device 100. Channel selection 156 may be done manually by the user or automatically. The motion waveform 158 may be viewed by the terminal device, for example in a suitable computing environment.

Algorithms may then be employed for automatic signal processing using machine learning. In particular, with reference to fig. 6A, a machine learning algorithm may be employed to implement at least the following two primary functions: automatic channel selection and bio-interface motion tracking.

Referring to the steps shown in fig. 6A, for channel selection, the RF signal may be scanned 162 and may be recorded 164 for a certain channel, which may then be converted 166 to an M-mode image. The image may be input into a developed Convolutional Neural Network (CNN) model. The predicted likelihood of the channel being in the correct position can be evaluated 168. After all channels 172 are scanned, the most likely channel may be determined or selected 174 and used for bio-interface motion monitoring. The peaks may be tracked 176 and a K-means clustering algorithm 178 may be used to identify 182 which portion of the signal represents the target biological interface. Finally, the motion of the target may be tracked by, for example, a Kalman filter, which is applied 184 to the identified signal regions.

Referring to fig. 6B, an illustration of a software design in accordance with the present principles, including autonomous artery identification and wall tracking, may be seen. The ultrasound RF data 175 produces a B-mode image 177 from which the subject can be located. This functionality can be achieved through various deep learning models designed for object localization. Continuous object tracking 179 may be performed by detecting objects through a series of successive frames, and wall tracking 181 using shifted signals, for example, may be performed by cross-correlation of the original RF signals (see fig. 7). Finally, the processed carotid artery wall waveform 183 may then be visualized on a graphical user interface.

As described above, when the interface moves, the reflection peak will move in the time domain corresponding to its motion. This can be seen in fig. 7, where the original peaks of the front and rear walls are shown as offset.

The whole system can integrate at least two functional modules: ultrasonic image enhancement, searching the position of the transducer so as to improve the quality of a reconstructed image; and an ultrasound image analysis that automatically analyzes ultrasound images acquired from the soft ultrasound probe.

With respect to the first major functional module, one major challenge in ultrasound imaging using a soft probe is that the location of the transducer elements is uncertain for most application scenarios. For proper reconstruction of the image, the transducer element positions should be determined with sub-wavelength level accuracy. In conventional ultrasound probes for diagnostic purposes, the transducer is fixed in a plane by a rigid housing. However, when integrated into the skin of a human body, the soft probe is positioned on and conforms to the dynamic curve surface and the position of the transducer will change constantly. Thus, if an appropriate method is not employed to compensate for transducer element displacement, the image reconstructed from the soft probe will be significantly distorted.

To address this problem, unsupervised machine learning algorithms can be applied to find the transducer locations, thereby improving the quality of the reconstructed image. The algorithm is inspired by the generation of the antagonistic network (GAN), as shown in fig. 8A. Fig. 8A shows the working principle and application of a conventional GAN, and in fig. 8B the proposed algorithm for ultrasound image quality enhancement is shown. The GAN is comprised of a generator 302 and a discriminator 304. The generator 302(G) synthesizes the images and the discriminator 304(D) attempts to distinguish these from a set of real images 303. These two modules are jointly trained so that D can only achieve random guess performance. This means that the G-synthesized image is indistinguishable from the real image. In the proposed solution, the GAN generator is replaced by a standard delay and sum (DAS) algorithm 305 for ultrasound image reconstruction, as shown in fig. 8B. Both modules may be trained using the large dataset from the ultrasound images 307 of commercial instruments as a training set for the real images. The algorithm takes the radio frequency voltage data acquired from the soft probe as input and learns the DAS beamformer parameters needed to reconstruct the ultrasound image. Training continues until these reconstructed images are indistinguishable from the existing real images.

With respect to ultrasound image analysis, a neural network-based model was developed to automatically analyze ultrasound images acquired from a soft ultrasound probe. A deep learning network trained for semantic segmentation may be used to extract blood pressure, blood flow, and heart pressure signals from the ultrasound images (M-mode 403, doppler 405, and B-mode 407, respectively). Typically, the model works well after training from a large image dataset. However, such data sets are unlikely to be used for soft probe ultrasound, at least in the initial phase. To overcome this problem, two sets of techniques are applied to support training of small data sets.

In more detail, fig. 9 illustrates a deep learning architecture (9A) and a bi-directional domain adaptation method (9B) for ultrasound image interpretation. Note that "EN" denotes an encoder network and "DN" denotes a decoder network.

The first technique of training with small data sets (as shown in fig. 9A) relies on parameter sharing between different tasks. This takes advantage of the fact that modern segmentation networks are implemented with codec pairs. The encoder abstracts the input image into a low-dimensional code that captures its semantic composition. The decoder then maps this code to a pixel-by-pixel partition. Typically, each task learns the network independently. However, this requires learning a large number of parameters. The architecture in this Al system includes the architecture shown on the right side of fig. 9A, where parameters are shared between tasks. In particular, the encoder 409 is shared by three tasks (411 and 413 and 415). Thus, the total number of parameters to be learned is reduced, suitable for training on small data sets.

Second, as shown in fig. 9B, relies on image transmission technology. The goal is to utilize the existing large ultrasound data set to help train the network of fig. 9A. The architecture here is domain adaptation. Domain adaptation a network trained on a large image dataset, in this example an existing ultrasound image, called the source domain, is applied to a new target domain, in this example a soft probe ultrasound image, in which the large dataset does not exist. This typically exceeds the performance of the network trained on the target domain. In this system, bi-directional adaptation is used to maintain the performance of the network. This iterates between the two steps. In a conversion step 421, an image-to-image conversion module 423 is used to convert the image of the existing ultrasound into an image of soft probe ultrasound. In an adaptation step 425, the segmentation module 427 trained on the former is transferred to the latter using a counterlearning procedure. The process iterates between two steps, gradually adapting to the network learned on soft probe ultrasound. The algorithm is applied to the architecture of fig. 9A to further increase the robustness of the segmentation.

Example (c): central blood pressure monitoring

In an exemplary embodiment, the systems and methods may be applied to a skin-integrated conformal ultrasound device 502 for non-invasively acquiring a Central Blood Pressure (CBP) waveform from a deeply buried blood vessel.

Figures 10A and 10B illustrate the use of a conformal ultrasound patch on a user. When mounted on the neck of a patient, the device may monitor the CBP waveform by transmitting ultrasound pulses to deep blood vessels. Fig. 10B illustrates the central blood vessel of the human neck. CA is the carotid artery, connected to the left heart. JV is the jugular vein connected to the right heart. Both arteries are located about 3-4 cm below the skin.

Because it is close to the heart, CBP can provide a better and more accurate method to diagnose and predict cardiovascular events than measuring peripheral blood pressure using cuffs. The conformal ultrasound patch can emit ultrasound which penetrates the human body by about 10 cm, measure the pulse wave velocity in the central blood vessel, and convert the pulse wave velocity into a CBP signal near the heart.

In addition, the blood pressure cuff can only determine two discrete blood pressure values, systolic and diastolic. However, blood pressure levels are dynamic every minute, fluctuating with our mood, excitement, meals, medications, and exercise. The cuff can only capture a snapshot of a segment. Since a conformal ultrasound patch can emit up to 5000 ultrasound pulses per second when worn continuously on the skin, a continuous beat-to-beat blood pressure waveform can be provided. Each feature in the waveform, such as a trough, notch, and peak, corresponds to a particular procedure in the central cardiovascular system, providing the clinician with rich key information.

As mentioned above, and as will be described in more detail below, the control electronics of the patch are able to focus and direct an ultrasound beam to accurately locate the target vessel, regardless of the location and orientation of the patch, so that any user error can be automatically corrected. The integrated bluetooth antenna may wirelessly transmit the blood pressure waveform to the cloud for further analysis.

In current clinical practice, CBP can only be achieved by implanting a catheter with a miniature pressure sensor into the vessel of interest. This type of measurement is typically performed in operating and intensive care units, is invasive and costly, and does not allow for routine and frequent measurements of the general population. Systems and methods in accordance with the present principles, using the described conformal ultrasound patch, not only results in improved diagnostic results and patient experience, but also gives the patient the ability to continuously self-monitor blood pressure anytime and anywhere. The large amount of data acquired can provide a basis for analyzing blood pressure fluctuation patterns, which is critical for accurate diagnosis and prevention of cardiovascular disease.

Fig. 11 shows an exemplary embodiment of a conformal ultrasound transducer array, which is indicative of conforming to a curved body surface.

Fig. 12-15 illustrate exemplary embodiments of systems and methods in accordance with the present principles, particularly dense array devices arranged for imaging and doppler ultrasound. In fig. 12, the reflected beam is received by the transducer array 102. To construct high resolution ultrasound images, dense arrays of transducers are often used. However, the dense arrangement of transducers sacrifices transducer size. Thus, each fine transducer element 116 within the array 102 will have a weaker signal amplitude compared to the large transducer.

To address this challenge, receive beamforming techniques have been developed. The ultrasonic signals received by each of the fine elements 116 are added to increase the signal-to-noise ratio according to the phase delay between the frequency channels. In other words, the raw signal 451 is calibrated to create a calibrated signal 453. In addition, receive apodization may be employed that uses a window function to weight the received signal (collectively referred to as steps and/or modules 455) to further enhance image contrast.

With this beamforming technique, non-destructive inspection of metal workpieces and biomedical B-mode images can be achieved using retractable ultrasound patches, as shown in the example application and shown in fig. 13A/13B and 14, respectively.

Transmission beamforming

Unlike conventional rigid ultrasound probes, which can easily create any desired doppler angle through probe manipulation, the retractable ultrasound patch cannot be physically tilted to create the proper angle of incidence for doppler measurements.

However, by utilizing transmit beamforming techniques, the ultrasound beam can be electronically tilted and focused. To achieve beam tilt and focus at the target point, especially on dynamic and complex curvatures, active and real-time delay profiles can be automatically calculated and applied to each transducer element. In particular, a real-time, high-speed phase aberration approach can be employed to accomplish this task. One of the main principles of phase distortion correction is that a signal received in one channel can be approximated by a delayed replica of the signal received in another channel. Thus, the time-of-flight error (i.e., phase distortion) is found in the cross-correlation function as the location of the maximum. Thus, the phase delay can be calculated to compensate for the error caused by the displacement of each unit. The beams emitted by each element interfere with each other to synthesize a highly directional ultrasound beam. By adjusting the determined time delay profile, the ultrasound beam can be tilted in a wide lateral window (from-20 ° to 20 °). The steerable ultrasound beam allows for the generation of an appropriate doppler angle at a particular organ/tissue of interest in the human body.

The following examples show the continuous monitoring of myocardial tissue contractility and carotid blood flow patterns, respectively.

In particular, fig. 16A and 16B show the application of the technique according to the prior art for tissue doppler imaging of myocardial tissue, and fig. 17A and 17B show the application of the technique according to the present principles, in particular for monitoring the blood flow in the carotid artery.

The system and method may be fully implemented in any number of computing devices. Generally, the instructions are disposed on a computer readable medium, typically non-transitory, and are sufficient to allow a processor in a computing device to implement the methods of the present invention. The computer readable medium may be a hard disk drive or solid state memory having instructions that are loaded into random access memory at runtime. Input to the application, for example from multiple users or from any one user, can be any number of suitable computer input devices. For example, a user may enter data related to a calculation using a keyboard, mouse, touch screen, joystick, touchpad, other pointing device, or any other such computer input device. Data may also be input through an inserted memory chip, hard disk drive, flash memory, optical media, magnetic media, or any other type of file storage media. The output may be delivered to the user by way of a video graphics card or integrated graphics chipset connected to a display that the user may see. Alternatively, a printer may be used to output a hard copy of the results. The present invention will also appreciate any number of other tangible outputs in view of this teaching. For example, the output may be stored on a memory chip, a hard disk drive, a flash drive, flash memory, optical media, magnetic media, or any other type of output. It should also be noted that the present invention may be implemented on any number of different types of computing devices, such as personal computers, laptop computers, notebook computers, netbook computers, handheld computers, personal digital assistants, mobile phones, smart phones, tablet computers, and devices specifically designed for these purposes. In one embodiment, users of smart phones or Wi-Fi connected devices download a copy of an application from a server to their device using a wireless internet connection. Appropriate authentication procedures and secure transaction processes may dictate payment to the seller. The application may be downloaded over a mobile connection, Wi-Fi, or other wireless network connection. The user may then run the application. Such networking systems may provide a suitable computing environment for an implementation method in which multiple users provide separate inputs to the system and method. In the following system, where patient monitoring is considered, multiple inputs may allow multiple users to enter relevant data simultaneously.

While the invention herein disclosed is capable of achieving the objects stated above, it is to be understood that this disclosure is merely illustrative of the presently preferred embodiments of the invention and that no limitations are intended other than as described in the appended claims. For example, the present invention may be used in a variety of environments.