The method for generating for identification and/or predicting the knowledge base of the failure of Medical Devices

文档序号:1757236 发布日期:2019-11-29 浏览:29次 中文

阅读说明:本技术 生成用于识别和/或预测医疗设备的故障的知识库的方法 (The method for generating for identification and/or predicting the knowledge base of the failure of Medical Devices ) 是由 莫里茨·于布勒 于 2019-05-16 设计创作,主要内容包括:本发明涉及一种生成用于识别和/或预测医疗设备的故障的知识库的方法,其中,医疗设备包括电气和/或机械部件以及医疗设备软件,其中,在医疗设备的操作中,医疗设备软件基于在医疗设备处测量的输出参数生成警告消息,其中,方法包括以下步骤:(a)提供虚拟部件,其中,每个虚拟部件对应于医疗设备的部件;(b)创建医疗设备的虚拟模型;(c)基于模拟至少一个虚拟部件的故障来呈现虚拟模型的至少一个虚拟输出参数;(d)将虚拟输出参数提供给医疗设备软件;(e)将医疗设备软件的响应与模拟故障相关联;其中,针对多个不同的模拟故障重复步骤(c)至(e),并且模拟故障和结果的相关联的响应用于建立医疗设备上的知识库。(The present invention relates to a kind of methods of the knowledge base of failure for generating and for identification and/or predicting Medical Devices, wherein, Medical Devices include electrical and/or mechanical part and Medical Devices software, wherein, in the operation of Medical Devices, Medical Devices software generates alert message based on the output parameter measured at Medical Devices, wherein, method is the following steps are included: (a) provides virtual component, wherein each virtual component corresponds to the component of Medical Devices;(b) dummy model of Medical Devices is created;(c) at least one virtual output parameter of dummy model is presented based on the failure for simulating at least one virtual component;(d) virtual output parameter is supplied to Medical Devices software;(e) response of Medical Devices software is associated with simulated failure;Wherein, step (c) to (e) is repeated for multiple and different simulated failures, and the associated response of simulated failure and result is for establishing the knowledge base on Medical Devices.)

1. method of the one kind for generating knowledge base (50), the knowledge base (50) is configured to identify and/or predict that medical treatment is set The failure of standby (10), wherein the Medical Devices (10) include electrical and/or mechanical part (11) and Medical Devices software (30), wherein in the operation of the Medical Devices (10), the Medical Devices software (30) is based in the Medical Devices (10) output parameter (12) measured at generates alert message (31), wherein the described method comprises the following steps:

(a) (101) virtual component (21) are provided, wherein each virtual component (21) corresponds to the portion of the Medical Devices (10) Part (11);

(b) dummy model (20) of (102) described Medical Devices (10) is created by using the virtual component (21);

(c) at least the one of (103) described dummy model (20) is presented based on the failure for simulating at least one virtual component (21) A virtual output parameter (22);

(d) Xiang Suoshu Medical Devices software provides at least one virtual output parameter (22) described in (104);

(e) response and described at least one by the Medical Devices software (30) at least one virtual output parameter (22) The failure of the simulation of a virtual component (21) is associated (105), and the response is especially raw by the Medical Devices software (30) At any alert message (31);

Wherein, step (c) to (e) is repeated for the failure of multiple and different simulations,

And wherein, the associated response of the failure of the simulation and the Medical Devices software (30) is for establishing the doctor Treat the knowledge base (50) on equipment (10).

2. according to the method described in claim 1, wherein, by the sequence of the dummy model (20) simulated failure, to generate The sequence of virtual output parameter (22) and corresponding alert message (31).

3. method according to one of the preceding claims, wherein consider the dummy section when simulating the failure The time-evolution of part (21).

4. method according to one of the preceding claims, wherein the knowledge base (50) includes machine learning system (40), especially deep learning system.

5. according to the method described in claim 4, further include:

The associated response of failure and the Medical Devices software (30) based on the simulation, especially by the police It accuses message (31), and is optionally based on the virtual output parameter (22a) of simulation, to train the machine learning system (40).

6. method according to claim 4 or 5, wherein the machine learning system (40) is by using the virtual mould At least one other virtual output parameter (22b) Lai Xunlian of type (20) simulation, the other virtual output parameter (22b) The output parameter (12) measured at the Medical Devices (10) in operation is not corresponded to.

7. the method according to any one of claim 4 to 5, wherein identified and mentioned by the machine learning system (40) For new output parameter (22b), wherein no with the output parameter (12) with measurement at the Medical Devices (10) in operation With mode measure the new output parameter.

8. the method according to any one of claim 4 to 6, wherein identified and mentioned by the machine learning system (40) For new service intervals, wherein the new service intervals are different from previous service intervals.

9. method according to one of the preceding claims, wherein the output parameter (12) of the measurement is sensor Output, voltage and or current.

10. a kind of computer program product, including program code instruction, said program code is instructed in said program code Instruction executes the step of method according to any one of claim 1 to 9 when being loaded into the memory of work station (60) Suddenly.

11. a kind of computer-readable medium is stored thereon with program code instruction, said program code instruction can be by work station (60) it reads and executees, to be executed according to claim 1 when executing said program code instruction by the work station (60) The step of to method described in any one of 9.

12. a kind of knowledge base (50) particularly includes machine learning system (40), the knowledge base (50) is by according to right It is required that method described in any one of 1 to 9 is established.

13. a kind of purposes of knowledge base according to claim 12 (50) is used to help Customer Service Center (35) as doctor It treats equipment (10) and safeguard service is provided.

14. according to the method for claim 13, wherein soft by Medical Devices in the operation of the Medical Devices (10) The alert message that part (30) generates is fed to the knowledge base (50), and in response, and the knowledge base provides probability value (51), wherein the probability value (51) specifies the probability of the failure of some component (11).

15. method described in 3 or 14 according to claim 1, wherein when by the subsequent alert message of Medical Devices software (30) generation (31) when, probability value (51) are adjusted by the knowledge base (50).

Technical field

The present invention relates to method, the computer journeys of the knowledge base for the failure for generating for identification and/or predicting Medical Devices Sequence product and computer-readable medium.The invention further relates to corresponding knowledge bases and such knowledge base to mention for Medical Devices For the purposes in customer service.

Background technique

In the prior art, Medical Devices especially medical imaging devices (such as magnetic resonance imaging (MRI) equipment or calculating Machine tomoscan (CT) imaging device) it is well known.These Medical Devices by such as patient table, X-ray tube, rack many portions Part therefore by the multiple mechanical of capacitor, resistor, circuit system, integrated circuit, microprocessor, memory module etc. and Electric component is constituted.These Medical Devices have high complexity.Many components are purchased from supplier, especially purchased from different Supplier, or developed by different team.For such highly complex equipment, only a small number of experts can be understood completely end and arrive The equipment at end.In addition, only collecting device specific experience can be carried out with the time and from available known device.

In order to guarantee the long-life of Medical Devices, usually provides and periodic maintenance is for example organized by Customer Service Center.Mainly Part is reaction equation service, wherein completing service after detecting problem by user.In addition, many Medical Devices are suitable for remotely Service, that is, Service Notification is triggered by Medical Devices and is for example sent to Customer Service Center via cloud.It is expected that being influenced in failure Such Service Notification is triggered before the operation of Medical Devices in time to provide maintenance well.These trigger events depend on The template developed based on the empirical value collected for certain types of Medical Devices.

However, such empirical value is not useable for Medical Devices newly developed, because required knowledge must pass through experience It collects at any time.Due to the high yield quality and stringent test of Medical Devices, it is greatly reduced in Medical Devices The shortage probability that breaks down when starting of service life.It therefore, only can be in the entire life of product of certain types of Medical Devices Period collects particular medical device when closing to an end (for example, when the Medical Devices of update further developed can be used) Empirical value.

Summary of the invention

The object of the present invention is to provide a kind of for supporting the operator for being responsible for maintenance to identify that Medical Devices (are especially newly opened The Medical Devices of hair) mistake or defect method.

It is (especially newly developed for predicting that Medical Devices may be influenced it is another object of the present invention to be provided for operator Medical Devices) operation defect device.

The purpose passes through method, the computer journey for generating for identification and/or predicting the knowledge base of failure of Medical Devices Sequence product, computer-readable computer medium, knowledge base and knowledge base purposes realize.

Processor below in relation to work station and solution according to the present invention is described about method claimed Scheme.Feature, advantage or alternative embodiment herein can distribute to other objects, and vice versa.It in other words, can be with Using in the described in the text up and down of method or feature claimed come the processor at improvement station.In this case, The functional character of method is embodied by the object element of processor.

According to the first aspect of the invention, a kind of method for generating knowledge base is disclosed, knowledge base is configured to know Other and/or prediction Medical Devices failures, wherein Medical Devices include electrical and/or mechanical part and Medical Devices software, Wherein, in the operation of Medical Devices, Medical Devices software generates warning based on the output parameter measured at Medical Devices Message, wherein method the following steps are included:

(a) virtual component is provided, wherein each virtual component corresponds to the component of Medical Devices;

(b) dummy model is created by using virtual component, wherein dummy model corresponds to Medical Devices;

(c) at least one virtual output ginseng of dummy model is presented based on the failure for simulating at least one virtual component Number;

(d) at least one described virtual output parameter is supplied to Medical Devices software, especially determines that Medical Devices are soft Response of the part at least one virtual output parameter, and

(e) by Medical Devices software to the response of at least one virtual output parameter (especially by Medical Devices it is soft Any alert message that part generates) it is associated with the failure of the simulation of at least one virtual component;

Wherein, step (c) to (e) is repeated for the failure of multiple and different simulations, and wherein, the failure of simulation and doctor The associated response of device software (30) is treated for establishing the knowledge base on Medical Devices.

It is compared with prior art, formed according to the present invention by alert message knowledge base associated with failure and defect, In, the experience of knowledge base is the result of simulation.By repeating step (c) to (e), if knowledge base is by simulating independent or dry part Failure or defect collect the experience about Medical Devices, rather than collect experience during the service life of actual medical equipment. Therefore, it can be advantageous to provide knowledge base when the service life of Medical Devices starts.It particularly, can be by based on warning Message identifies possible failure to support the operator for being responsible for maintenance, this is because knowledge base makes operator's understanding may be with warning The associated latent defect of message.For example, several potential failures or defect may cause identical alert message, and know Know library and advantageously provide the idea for being most likely to be which latent defect, especially disappears at the age of Medical Devices or another warning Under the background of breath.Present invention additionally contemplates that Medical Devices software is usually infrequently modified, therefore, Medical Devices software cannot hold It changes places the newly identified correlation being updated to reflect that between defect and output parameter.Therefore, operator depends on Medical Devices The content of the response of software, especially alert message.By using the present invention, it advantageously may rely on knowledge base and come according to police It accuses message identification or isolation defect, the correlation especially between defect and output parameter is indefinite under such circumstances.

" knowledge base " may include that the relevant response of many simulated failures and Medical Devices software (especially corresponds to warning to disappear Breath interior the fact that do not give a warning message perhaps) set.In its simplest form, knowledge base be include such as mould The database of the list of the associated analog response of quasi- failure and Medical Devices software.In this case, " correlation " is knowledge Following record in library: certain response of Medical Devices software has occurred in the result as specific fault.In other embodiments In, the correlation between alert message and defect considers the other information collected in the simulation of failure, such as in the simulation phase Between " measurement " one or several virtual output parameters.In useful embodiment, knowledge base is trained to consider such as Several aspects at the age of Medical Devices and other previous alert messages, to be used for one or more defects and a spy Fixed alert message is associated.In other words, for knowledge base instead of the protracted experience of operator, operator is, for example, that can pass through The mankind client service person of defect is identified according to general alert message using his experience in the previous service life of Medical Devices Work.Therefore, the present invention is particularly useful in terms of Medical Devices newly developed not available for protracted experience provide safeguard service 's.Preferably, knowledge base provides the probability and/or different incipient faults and corresponding probability value of specific failure or defect List.Particularly, knowledge base includes the table of correlation and/or the machine learning algorithm based on correlation training.

In the preferred embodiment, term " output parameter " indicate can during the operation of actual medical vision facilities The measurable parameter detected at actual medical vision facilities.For example, output parameter is the voltage etc. of particular comparator.Output ginseng Number can be the frequency and/or amplitude of signal.Signal can be number or complicated, such as the error message of individual component And/or error code.In general, the deviation of output parameter be originated from defect, the defect with offer output parameter component and/or and mention For output parameter component cause in relation to and due to its defect and default output parameter deviation different components it is associated. For example, capacitor has defect, and/or the resistor for being connected to capacitor has defect.The two defects all may cause with The deviation of output parameter.In addition, " virtual output parameter " corresponds to " output parameter " measured in operation, but it is simulation As a result.Therefore, can by include its simulation virtual output parameter dummy model come replace Medical Devices and with when Between the output parameter of Medical Devices that measures.

Term " alert message " is in this application for any kind of notice (especially in the biography for being provided to operator The form for the information sent and shown), and the information including the state about Medical Devices.Alert message is by Medical Devices software It generates the response as the output parameter to measurement and is for example shown on the screen, or for example especially can automatically pass through The Customer Service Center of a part as long distance service system is sent to by cloud or by internet.Therefore, useful In embodiment, Medical Devices software control and/or the function of monitoring Medical Devices.Alert message is forwarded or is sent to service Then device and shows message especially to the central management server for for example coordinating maintenance on the screen.Preferably, knowledge Warehouse compartment is at server.For example, after identifying the exception of output parameter, directly to operator on the screen of Medical Devices Show alert message.Therefore, operator can adjust further maintenance strategy in real time.Particularly, alert message include with Lower information: the difference between standard value and reality output parameter is more than threshold value.For example, alert message points out " the capacitor B on mainboard The voltage at place drops to 0.5 μ V " of value.In this embodiment, alert message is to the failure or defect of operator notification component, And it is to provide the information of the deviation about specific or several different specific output parameters and corresponding default value.Such warning disappears Breath is obviously the most useful in terms of providing safeguard service, because such alert message allows to provide following service: for example existing Medical Devices are entirely ineffective and cause to replace defective component before shutting down.

Term " dummy model " indicates the design of simulation Medical Devices, the especially arrangement and/or connection of component.Especially Ground, " dummy model " are formed by virtual component corresponding with the actual part of Medical Devices.For example, using such as Agilent The Visual Programming software of VEE or Labview creates the dummy models of medical imaging devices.In such Visual Programming software work In tool, each virtual component can by and circuit diagram it is comparable in a manner of combine and be connected to each other.Particularly, dummy model is independent Consider effect caused by the interaction between the effect or influence and/or component of each component.For example, can be known that The defects of one component causes the specified defect being connected in the second component of the first component.Even if the defect is to second component Output does not influence, and simulation is it is also contemplated that the effect, because of its function that may influence to be connected to the third member of second component Energy.It is further contemplated that third member is connected to the first component and second component.Preferably, then, first is depended on The failure of the failure of part, the failure of second component and the first component and second component simulates the effect difference to third member, Wherein, to the effect of the failure of the first component and second component with individually as caused by the failure of the first component and second component The addition of effect is different.Particularly, it distributes to corresponding component or the interaction generation by component effect and is included in corresponding component Failure simulation in.In embodiments, by operator for example by using the man-machine of keyboard, mouse, touch tablet etc. Interface is arranged dummy model.In other embodiments, the number according to used in the construction in certain types of Medical Devices Word construction plan (especially CAD (CAD) file) Lai Shengcheng dummy model.Such cad file particularly wraps The electrical layout of type containing Medical Devices, and include a part of mechanical realization or mechanical realization.Particularly, virtual mould Type can be " the digital twins " of Medical Devices.In general, dummy model will be to the manufacture in many individual samples and sale Certain types of Medical Devices modeled.Dummy model is simulated as caused by the failure or failure of one or several components All subsequent affects.These influences may cause the exception of one or several output parameters.For example, main capacitor is in detector It breaks down on circuit board, this will lead to voltage decline etc..Finally, this will lead to the failure of some equipment.In other words, mould It is quasi- that each voltage and each successor are modeled and output data is presented to be supplied to actual Medical Devices software.

Preferably, the term used in the context of the generation of knowledge base " Medical Devices " describes certain types of doctor Treat equipment, especially medical imaging devices, such as a type of nuclear medicine (NM) scanner, positron emission computerized tomography (PET) scanner, X-ray scanners (XA), fluoroscopy (RF) scanner or breast imaging (mammography) (MG) scanning Instrument.For example, the type of Medical Devices is classified by its version or generation.However, knowledge base is preferably used for helping to safeguard the spy One or several the individual Medical Devices for determining type, be particularly wall-mounted at medical centre and for its provide (such as Remotely) the individual Medical Devices of customer service.For example, providing the risk of failure for individual Medical Devices.

As disclosed in as described below, especially advantageous embodiment of the invention is provided by dependent claims And feature.The feature of different claim categories can be appropriately combined to provide other embodiments not described herein.

Therefore, the invention further relates to be configured to identify and/or predict Medical Devices failure knowledge base purposes, In, Medical Devices include electrical and/or mechanical part and Medical Devices software, wherein in the operation of Medical Devices, medical treatment Device software generates alert message based on the output parameter measured at Medical Devices, wherein passes through what is included the following steps Method generates knowledge base:

(a) virtual component is provided, wherein each virtual component corresponds to the component of Medical Devices;

(b) dummy model is created by using virtual component, wherein dummy model corresponds to Medical Devices;

(c) at least one virtual output ginseng of dummy model is presented based on the failure for simulating at least one virtual component Number;

(d) at least one described virtual output parameter is provided to Medical Devices software;

(e) by Medical Devices software to the response of at least one virtual output parameter (especially by Medical Devices it is soft Any alert message that part generates) it is associated with the simulated failure of at least one virtual component;Wherein, for multiple and different Simulated failure repeat step (c) to (e), and wherein, the simulated failure and associated response of Medical Devices software (30) For establishing the knowledge base on Medical Devices.

According to useful embodiment, during the operation of individual Medical Devices, i.e., after having built up knowledge base, Knowledge base can will be fed to by the alert message of Medical Devices Software Create, and provide probability value by knowledge base, wherein probability Value specifies the probability of the failure of some component.Therefore, cause the probability of alert message by specified defect to operator notification.Cause This, for example, operator can assess whether to be immediately required to what maintenance or whether repair in periodic maintenance next time was identified The failure of component is enough.Particularly, list of probabilities can be provided, wherein list respectively carries out the probability of specified defect Sequence.Be read for example, being listed as follows as caused by alerting: " defect A causes the probability of alert message to be a%, and defect B causes The probability of alert message is b% ".Therefore, several latent defects and corresponding probability are provided to operator.Such probability Value (can analyze the failure for causing alarm specific message that is, passing through by using the statistical estimation to simulated failure by knowledge base And correspondingly allocation probability value) Lai Shengcheng.In another useful embodiment, knowledge base includes machine learning system, should Machine learning system using such as deep learning algorithm and accordingly by simulated failure be trained to according to alert message come Identify some failure.

According to useful embodiment, it is assumed that adjusted when alert message subsequent by Medical Devices Software Create by knowledge base Probability value.For example, knowledge base has following information: combination of two different alert messages in some time interval indicates certain The specified defect of a component at least increases the probability that defect occurs.Therefore, it can be extracted in addition from the sequence of alert message Information.For example, above-mentioned be listed as follows is read: " probability that defect A occurs is changed to a'% from a%;Defect B occurs general Rate is changed to b'% " from b%.

Preferably, it is assumed that by dummy model come the sequence of simulated failure for generating virtual output parameter and corresponding police Accuse the sequence of message.Therefore, it can be advantageous to which alert message sequence other is lacked with may originate from causing in other component The failure sequence of a sunken component individually to fail is associated.Therefore, the alert message after first alert message can be with For defective FRU to be isolated.Furthermore, it is possible to it is contemplated that, the time develops, i.e. time interval between two alert messages, It can be used for identifying defective one or more components.Particularly, term " sequence " expression generates several one by one Alert message.Particularly, the type of the period between two alert messages and/or the alert message followed each other can be used for knowing Other or prediction failure.For example, the first alert message points out " voltage of condenser B has declined ".Subsequently generate second message simultaneously And second message points out " voltage of condenser C has declined ".The sequence of the two alert messages can indicate that medical imaging is set Some standby defect or failure are possibly even different from reasonable defect possible after receiving the first alert message.It is excellent Selection of land, it is also contemplated that the time interval between the first alert message and the second alert message, for excluding latent defect or for pre- Survey failure.

Particularly, it is assumed that consider the time-evolution of virtual component in simulations.This means that consider the old of virtual component Change.For example, the probability that component is reacted in some way depends on its age.The aging of component may be it is known and by It is integrated in simulation, so that virtual component also depends on the simulation year of virtual component to the reaction of the defect of another virtual component Age.Therefore, it can be advantageous to consider the aging of each component.This means that the timing of alert message can be used (i.e. in knowledge base Since being installed particular medical device until alert message occur time interval) come identify specified defect with true medicine at As equipment operation in the associated probability of alert message that provides.

According to the preferred embodiment of the present invention, it is assumed that knowledge base includes machine learning system, preferably by depth Learning system.In some embodiments, knowledge base is machine learning system.Particularly, by knowledge base input (for example, Alert message and simulated failure) and the virtual output parameter of optional simulation carry out training machine learning system.By using machine Device learning system, it can be advantageous to identify unconspicuous universal when simply arranging the data of collection in the database And/or correlation.Particularly, machine learning system uses the artificial network realized for example at server.In addition, machine learning System will can seem at first glance that incoherent alert message is associated each other.Particularly, it is assumed that housebroken artificial Network has input layer, multiple hidden layers and output layer.For example, output parameter and/or alert message are provided to input layer, and And preferably, each potential output parameter and/or each potential alert message are assigned to the node of input layer.Warning disappears The sequence of breath is also used as input parameter, then advantageously, the sequence of potential output parameter and/or potential alert message It is assigned to the node of input layer, in addition it is possible to use time difference between the measurement of potential output parameter and/or potential Time difference between alert message is used as input parameter, particularly to assign these to the node of input layer.Particularly, it inputs The node of layer can correspond to two-dimentional input, wherein the first dimension corresponds to potential output parameter and/or potential warning disappears Breath, and wherein, the second dimension corresponds to time interval.Then, output layer provides some component tool defective probability.Output Each node of layer is assigned to some component, and exports the correspondence probability of the defect of some component.In order to train artificial net Virtual output parameter and the information about corresponding defect are supplied to artificial network by network.Particularly, virtual output parameter is mentioned Supply the node of input layer.In addition, training artificial network by the sequence of the virtual output parameter generated by dummy model.Cause This, knowledge base is based on virtual output parameter and preferably in their order come the experience of collecting.

If machine learning system helps in view of the failure for simulating each component and/or dry part can cause mass data It, can the information used in the knowledge base of the Medical Devices of the type to extract in the so big quantity of processing.Engineering Learning system is by knowledge base from defect and the associated knowledge base of alert message are only converted into veteran knowledge base.It is another excellent Point is: machine learning system realizes correlation and correlation so that after the small hardware at the Medical Devices changes or Person does not need to repeat entire simulation in the case where the small software about Medical Devices software changes.In addition, machine learning system System not only can be by being trained visible (simulation) alert message of operator, can also be by being provided by dummy model Virtual output parameter be trained, the virtual output parameter is usually fed to Medical Devices software.Due to machine learning System can handle more information, therefore can also provide such depth system information of Medical Devices.

In useful embodiment, come by using at least one other virtual output parameter of dummy model simulation Training machine learning system, the other virtual output parameter do not correspond to the output ginseng measured at Medical Devices in operation Number.Since machine learning system is not instead of from actual medical learning equipment, learn from its dummy model, therefore be not limited to existing Grow directly from seeds it is living in such Medical Devices at handle actually available (that is, actual measurement) output parameter (for example, voltage, electricity Stream ...).But dummy model can provide various output parameters;It " can measure " be practically without such sensor can Voltage or electric current at point.This will improve the accuracy for any prediction done by machine learning system.

In another embodiment, it is assumed that identified by machine learning system and new output parameter is provided, wherein with The different mode of the output parameter measured at Medical Devices in operation measures new output parameter.Therefore, machine learning system It can determine the optimization measurement point in Medical Devices.For example, machine learning system suggestion measures another output parameter, such as different Voltage at capacitor because the new output parameter of measurement provide specified defect for identification or with greater number of component phase Associated improved reliability.It is further contemplated that machine learning system suggestion measures new output parameter and is used as to spy Determine the response of alert message.For example, machine learning system identifies two latent defects associated with an alert message, and Machine learning system, which is also realized that measure another new output parameter, can exclude or confirm one in these latent defects.Cause This, machine learning system can suggest measuring new output parameter.It is even contemplated that, knowledge base disappears in the warning Breath notifies operator to measure new output parameter after being generated, to accelerate the identification to defect.

Moreover it is preferred that assuming to be identified by machine learning system and provide new service intervals, wherein between new maintenance Every different from previous service intervals.For example, machine learning system is in one or several alert messages instruction Medical Devices Suggest changing service intervals when component or subelement break down in the near future.Hence, it can be determined that being tieed up for predictability Protect the optimization sampling period in useful Medical Devices.

In order to provide new output parameter and/or new maintenance model, for different output parameters and/or different polices Message sequence is accused to simulate dummy model.Therefore, the output parameter of simulation can be supplied to artificial network, be especially available to Certain nodes in input layer, and check whether the probability that some defect can be inferred to according to new output parameter.For example, Compared with previous output parameter, new output parameter reduces the number of latent defect.It is further contemplated that trained Virtual artificial network predict the key of a component according to having the alert message of alert message sequence in the near future Defect.By adjusting maintenance timing, can make a response in time.

In another embodiment, it is assumed that the output parameter of measurement is the output of sensor, especially electric current, voltage.Also Sensor can be used to observe the state of mechanical part.For example, the vibration for the magnet that sensor detection is used by magnetic imaging equipment It swings.For example, the change of frequency may be the result of the defect at magnet.Furthermore, it is possible to it is contemplated that, continuously or at intervals Observe output parameter.Particularly, it is assumed that the generation of the deviation trigger alert message of the parameter and default value of measurement.Other component It can be patient table, X-ray tube, capacitor, resistor, circuit system, integrated circuit, microprocessor, memory module etc..

In another embodiment of the present invention, it is assumed that knowledge base is used to design the Medical Devices of new version.Particularly, know Knowledge library can identify the design defect under the early stage state of newly-designed Medical Devices.

In addition, knowledge base can be used for optimizing data collection, to save data bandwidth, because machine learning system can be with It is recommended that suitable output parameter and the interval that these parameters should be measured and possible service intervals, to more efficiently carry out Maintenance and service.

Another aspect of the present invention is a kind of computer program product comprising program element, the program element are used for The step of being executed according to the method for the present invention when program element is loaded into the memory of work station.

Another aspect of the present invention is a kind of computer-readable medium, is stored thereon with program element, described program element It can be read and executed by work station, so as to the step of program element is performed execution according to the method for the present invention.It calculates Machine readable medium can be the data medium of such as cloud, hard disk, CD-ROM, SD card or other digital medias.Work station can be Any type of digital calculating equipment, for example, PC, smart phone, tablet computer, cloud, work station or handheld device.It is preferred that Ground, work station include processor described below.

Preferably, another aspect of the present invention is the processor of work station, which is configured for executing following At least one of step:

There is provided virtual component, wherein each virtual component corresponds to the component of Medical Devices;

Dummy model is created by using virtual component, wherein dummy model corresponds to Medical Devices;

The virtual output parameter of dummy model is presented based on the failure for simulating at least one virtual component;

Virtual output parameter is supplied to Medical Devices software;

By Medical Devices software to the response of virtual output parameter (especially by any warning of Medical Devices Software Create Message) it is associated with the simulated failure of at least one virtual component;

For multiple and different simulated failures, previous at least some of step is repeated, and/or

Establish the knowledge base on Medical Devices.

Term " work station " preferably refers to (individual) computer, the virtual machine run on host hardware, microcontroller Or integrated circuit.As alternative, work station can be true or virtual calculating unit (the true skill for calculating unit Art term is " cluster ", and the technical term of virtual calculating unit is " cloud ").Preferably, work station includes computing unit, processing Device, interface and/or memory cell.Computing unit may include hardware element and software element, such as microprocessor or scene Programmable gate array.Memory cell can be presented as impermanent main memory (such as random access memory) or permanent great Rong It measures storage device (such as hard disk, USB stick, SD card, solid-state disk).Work station can be a part of Medical Devices.

The invention further relates to the knowledge bases established by means of the present invention.Particularly, the present invention relates to through the invention Method training machine learning system.Such machine learning system provides useful help, client to Customer Service Center Service centre provides safeguard service to the Medical Devices for the type simulated in dummy model.

Therefore, the invention further relates to knowledge bases (preferably include machine learning system or be made of machine learning system) Purposes is used to help Customer Service Center as Medical Devices and provides safeguard service.

From anywhere in not being expressly recited, each embodiment or their own aspect and feature can groups each other Close or exchange, without limiting or expanding described the scope of the present invention, so long as combination or exchange it is meaningful simultaneously And in the sense of the present invention.Under applicable circumstances, the advantages of describing about an embodiment of the invention is also this hair The advantages of bright other embodiments.

Detailed description of the invention

Illustrated the present invention by means of specific embodiment now with reference to attached drawing, in the accompanying drawings:

Fig. 1 schematically shows the side for showing the failure for predicting Medical Devices according to another preferred embodiment The block diagram of method;

Fig. 2 schematically shows the examples of the simulation for the dummy model for using Fig. 1;

Fig. 3 schematically shows the example of knowledge base;And

Fig. 4 schematically shows the sides for showing the failure for predicting Medical Devices according to another preferred embodiment The flow chart of method.

Specific embodiment

In Fig. 1, it is schematically shown that the block diagram of the method for the latent defect of Medical Devices 10 for identification is shown.Figure The relevant action that 1 upper part 58 shows individual Medical Devices 10 and occurs in the operation of such equipment, and The low portion 61 of Fig. 1 shows dummy model and relevant method and step.

Medical Devices 10 can be medical imaging devices, MR imaging apparatus or ct apparatus.These Equipment 10 is formed by multiple components 11, is preferably formed by mechanically and electrically component 11.Each of component 11 can be can It can lead to the error source of the failure of Medical Devices.In order to reduce the probability of failure, it is periodically executed maintenance process.For example, safeguarding Journey is triggered by Medical Devices software 30, which is based on being existed by sensor 14 during the operation of Medical Devices 10 The output parameter 12 (for example, voltage of electric component) measured at Medical Devices 10 generates alert message 31a.The output of measurement is joined Number 12 is used as the input of Medical Devices software 30, to identify the exception and/or defect of Medical Devices 10.For this purpose, medical treatment is set Standby 10 include the analytical unit for executing Medical Devices software 30.For example, these alert messages are sent to cloud, network or clothes Business device, especially service management system, to notify the exception detected at Medical Devices 10 to operator 35.It can be with structure It is contemplated that Medical Devices software 30 is located at server, and output parameter 12 is sent to server.Particularly, work as knowledge Not Chu output parameter 12 significant modification when, Medical Devices software 30 initiate or triggering alert message 31 communication.It can be with structure It is contemplated that output parameter 12 specifies the state of mechanical part 11 by using corresponding sensor 14, such as stage body is determined To.Either directly by Medical Devices 10 still in service at the heart, alert message 31a can be shown at screen 32 To operator 35.

In general, alert message 31 notifies the particular state of output parameter 12 to operator 35.For example, alert message 31a table Show that " voltage of capacitor drops to critical level." however, the reason of particular state may be various.For example, voltage Decline is as caused by the defect of resistor, or is caused by the defect as another component 11 as an alternative.Therefore, it is responsible for maintenance Operator defect or problem can not be distributed to clearly to the content of alert message 31.However, such equipment 10 have through The expert tested can be identified or be associated with defect due to its experience based on the content of alert message.

Due to the quality of component 11, the risk or probability of mistake or defect are very low, especially in corresponding Medical Devices 10 When service life starts.Therefore, almost without empirical value can be used for based on the alert message 31 provided by Medical Devices software 30 come Identify this defect or latent defect.Defect or mistake still possibly even occurs when next-generation Medical Devices 10 can be used. It cannot function as empirical value for Medical Devices 10 of new generation accordingly, with respect to the information of these defects.

In order to optimize the service life of Medical Devices 10 and support operator by identification mistake, it is assumed that creation or adjustment medical treatment The dummy model 20 of equipment 10, the dummy model 20 of especially individual Medical Devices 10.Preferably, including the work of processor Stand 60 be configured for create and provide Medical Devices 10 dummy model 20.Therefore, processor is preferably incorporated in work It stands, in (individual) computer, server and/or cloud computer.For example, being that void is separately provided in particular medical device 10 by operator Analog model.Therefore, dummy model 20 corresponds to Medical Devices 10.Particularly, dummy model 20 is by the true portion with Medical Devices The formation of the corresponding virtual component 21 of part 11.Therefore, dummy model 20 considers effect and itself and other portions of corresponding component 11 The interaction of part 11.

It is also contemplated that the construction plan of Medical Devices 10 is for example provided to processor in the form of cad file, It identifies each component 11, and constructs dummy model 20.For example, scanning construction plan, and processor is based on the structure from scanning It makes the information extracted in the works and constructs dummy model 20 automatically.

In order to identify potential mistake and failure, it is assumed that by simulating virtual component 21 (especially each virtual component 21) Failure or defect and these virtual components 21 failure each combination, come present respectively with the progress at point 24a and point 24b Virtual measurement corresponding virtual output parameter 22a, 22b.Obtained virtual output parameter 22a, which is provided to, to be for example mounted The copy of Medical Devices software 30 on processor or network.It therefore, can be by the response and simulation of Medical Devices software 30 Failure it is associated.By being repeated for multiple and different simulated failures to this, knowledge base 50 advantageously can establish.It changes Sentence is talked about, in order to collect the experience about Medical Devices software 30 to the response of potential output parameter 12, by dummy model 20, Medical Devices 10 are replaced especially by the virtual output parameter 22 simulated for individual Medical Devices 10.

In dummy model 20, it not only may collect at point 24a corresponding with the sensor 14 in Medical Devices 10 and survey The virtual output parameter 22a of amount.Furthermore, it is possible to collect additional virtual output parameter, such as from actual medical devices 10 The output parameter 22b for not having the point 24b of respective sensor to obtain.However, such virtual output parameter 22b is being collected about doctor It is useful when treating the deep knowledge of equipment 10, therefore is provided to knowledge base 50.In addition, knowledge base 50 is it can be found that such Additional virtual output parameter 22b is more useful when providing the information about the present or future defect of Medical Devices, and therefore Can suggest should at point 24b rather than measured at 24a, and therefore can design future type Medical Devices When useful input is provided.

Preferably, the machine learning system 40 of such as deep learning system is used as a part of knowledge base 50, especially Correlation between the failure/defect simulated for identification and alert message 31b.Such machine learning system 40 preferably collects It is trained in network or server, and through (simulation) alert message 31b and the defect of simulation, and optionally It is trained by corresponding virtual output parameter 22a and 22b.

Therefore, the network including machine learning system becomes the expert about individual Medical Devices 10.For example, engineering Learning system 40 identifies defect and/or trained based on the sequence of the alert message 31a generated from independent Medical Devices 10 Knowledge base 50 specified defect has been occurred according to alert message 31a offer or will occur the probability of further disadvantage.In addition, machine Device learning system 40 can be provided to the operator 35 of Medical Devices 10 occurs entirely ineffective probability in limiting time interval 51.It therefore, can effectively organizational protection.

It is further contemplated that being configured for identification than by default or past measurement to, machine learning system 40 Those output parameters 12 are more suitable for isolation/identification defect output parameter.By changing output parameter 12, can correspondingly adjust Whole Medical Devices software.It is further contemplated that machine learning system 40 determines the timing for measuring output parameter 12.Cause This, can optimize the sampling to output parameter 12, so that reduce the amount for being sent to the alert message 31a of network/server, And improve the quality for the appropriate message for including in alert message 31a.

In fig. 2 it is shown that the example of the simulation using the dummy model 20 of Fig. 1.In this example, Medical Devices 10 wrap Include such as gate drivers driver element 18 and first component 11', second component 11 " and third member 11 " '.In order to Simulate the failure in corresponding dummy model 20, it is assumed that driver element 18 has specified defect, such as the short circuit of capacity.Therefore, The normal output that the defect of virtual component corresponding with driver element 18 exports 18 " and usually exported by driver element 18 18' is different.However, defective output 18 " causes the decline of the inductance at virtual component corresponding with first component 11', example The decline of voltage such as from 0.78 μ H at 0.1 μ H, and virtual component corresponding with second component 11 ", for example, from 1.2V to 0.2V is (in Fig. 2, respectively at the top of each component: the normal expected value of output 18';In bottom: defective output 18 " Expected value).Changing for described parameter simulation can be determined by the failure in simulation dummy model 20.It can be with structure It is contemplated that defective output 18 " is with the component of third member 11 " ' corresponding virtual component on not influencing such as, this It can be determined by the failure in simulation dummy model 20.In shown simulation, resistance keeps 0.1 Ω.However, with Defective output 18 " at the corresponding virtual component of driver element 18 is to the virtual output parameter 22a measured at point 24 Have an impact.Therefore, it by the failure in simulation dummy model 20, can be generated all as caused by some defective output 18 " Such as the virtual output parameter 22a of output voltage, which drawn by some component of such as driver element 18 It rises.

In fig. 3 it is shown that the example of knowledge base 50.For example, defect A, knowledge base 50 (i.e. lacks three different defects Fall into B and defect C) distribute to alert message 1.In addition, knowledge base 50 specifies the correspondence probability of individual defect.Probability is specified by right Defect is answered to cause the probability of alert message.For example, knowledge base 50 is by probability A' and defect A, probability B' and defect B and probability C' It is associated with defect C, wherein probability A' is the probability for causing alert message 1 by defect A, wherein probability B' is drawn by defect B The probability of alert message 1 is played, and wherein, probability C' is the probability for causing alert message 1 by defect C.It is further contemplated that arrive It is that knowledge base 50 only identifies typical defect, such as defect D for alert message 2 in the case where not specified probability and lacks Fall into F.In some cases, alert message 3 is possibly even only associated with a latent defect.Particularly, in addition, defect can be with Specify a defective component or several components.

In fig. 4 it is shown that the method for showing the failure for predicting Medical Devices according to another preferred embodiment Flow chart.Particularly, method includes:

(a) 101 virtual components are provided, wherein each virtual component corresponds to the component of Medical Devices;

(b) dummy model 20 of 102 Medical Devices 10 is created by using virtual component 21;

(c) at least one for being presented 103 dummy models 20 based on the failure for simulating at least one virtual component 21 is virtual Output parameter 22;

(d) 104 at least one virtual output parameter 22 are provided to Medical Devices software;

(e) response by Medical Devices software 30 at least one virtual output parameter 22, it is especially soft by Medical Devices Any alert message 31 that part generates, associated with the simulated failure of at least one virtual component 21 105;

(f) step (c) to (e) is repeated for multiple and different simulated failures, and

The alert message generated by Medical Devices software 30 is fed to knowledge base 50, and in response, knowledge base 50 There is provided probability value 51, wherein probability value 51 specifies the probability of the failure of some component 11.

16页详细技术资料下载
上一篇:一种医用注射器针头装配设备
下一篇:一种用于CRRT的远程监测系统和设备

网友询问留言

已有0条留言

还没有人留言评论。精彩留言会获得点赞!

精彩留言,会给你点赞!