Medical device failure prediction

Abstract

Proposed are concepts that assist in fault resolution by identifying new sets of factors that may contribute to the occurrence of one or more faults. In particular, embodiments of the invention propose to predict or derive unforeseen faults from existing or known faults based on associated parameters, events, etc. (i.e., contributing factors). By identifying unforeseen / unrecorded faults, a solution can be identified prior to the occurrence of a problem, potentially reducing the solution / repair time, cost, and downtime of the medical device.

Description

Technical Field

[0001] The present invention relates to the field of equipment maintenance, and in particular to the field of predicting failure of medical equipment. Background Art

[0002] Medical imaging equipment may occasionally malfunction, i.e., exhibit a malfunction and fail to operate properly.

[0003] Typically, in response to a fault in a medical imaging device, one or more service engineers are assigned to diagnose and resolve the fault. Such service engineers may perform the diagnosis remotely or on-site.

[0004] In such a workflow, it is not always possible to list all problems (problems / issues) that may occur in a medical system during development and prepare corresponding solutions. When such unforeseen problems occur, a lot of time and effort from service engineers is required to diagnose and solve the problems.

[0005] Therefore, it is desirable to reduce the length of time it takes to identify the cause of, or a solution to, a malfunction in a medical imaging device.

[0006] Document US2019 / 357873A1 discloses a method for generating a knowledge base for identifying and / or predicting failures of a medical device. The method comprises: a) providing virtual components, each corresponding to a component of the medical device; b) creating a virtual model of the medical device using the virtual components; c) presenting at least one virtual output parameter of the virtual model based on a simulation of a failure of at least one of the virtual components; d) providing the presented at least one virtual output parameter to the medical device software; e) associating a response of the medical device software with the presented at least one virtual output parameter; steps c) to e) are repeated based on a plurality of different simulated failures, and the plurality of different simulated failures and the associated responses of the medical device software are used to generate the knowledge base on the medical device.

[0007] Document JP2022022497 A1 discloses a fault diagnosis system, which includes: a communication unit that performs predetermined communication with a terminal; a fault causal model that includes causal relationships between multiple fault modes and causal relationships between multiple inspection items and multiple fault modes; a fault mode probability estimation unit that calculates the occurrence probability of each fault mode in multiple fault modes based on the results of multiple inspection items received via the communication unit; and a terminal display information generation processing unit that generates information for displaying the estimation result of the fault mode probability estimation unit 24 associated with the fault causal model on the terminal. Summary of the Invention

[0008] The invention is defined by the claims.

[0009] According to an example according to one aspect of the present invention, there is provided a method for predicting factors contributing to a failure of a medical device, the method comprising:

[0010] obtaining a first set of contributing factors leading to a first failure of the medical device;

[0011] inputting the description of the first fault of the medical device and the first set of contributing factors into a machine learning algorithm, the machine learning algorithm comprising a generative neural network trained to predict a predicted second set of contributing factors that will lead to occurrence of the first fault of the medical device based on the input description of the first fault and the first set of contributing factors, the second set of contributing factors being different from the first set of contributing factors, and wherein the generative neural network is trained to predict the second set of contributing factors using training data comprising a plurality of fault descriptions and, for each fault, a corresponding set of contributing factors that led to the fault;

[0012] generating a predicted second set of contributing factors using the trained machine learning algorithm in response to the input; and

[0013] The second group of contributing factors is output.

[0014] Therefore, the proposed concepts are intended to provide schemes, solutions, concepts, designs, methods and systems related to assisting fault resolution by identifying a new set of factors that may contribute to the occurrence of one or more faults. In this way, unforeseen / undocumented problems / faults that may occur during the use of medical equipment / devices can be identified a priori.

[0015] Various embodiments can predict or derive unforeseen faults based on associated (log file) parameters, events, etc. (i.e., contributing factors) from existing or known faults. Examples of associated log file parameters include values for settings used to control medical devices, as well as values for monitoring current, voltage, temperature, speed, etc., monitored during operation of the medical device. The values of the log file parameters can be extracted from the log files of the medical device. By identifying unforeseen / undocumented problems / faults, solutions can be identified before problems occur, potentially reducing resolution / repair time, costs, and downtime of the medical equipment.

[0016] The proposed concept may also enhance reliability during the development phase of medical devices by providing information about failures and associated solutions, which can be utilized to help avoid or prevent failures during the device use phase.

[0017] Specifically, embodiments of the present invention provide for using a first set of contributing factors that led to a first failure of a medical device to generate (e.g., predict) a second, different set of contributing factors that will lead to the occurrence of the first failure. The second, different set of contributing factors can be used to identify one or more new failures that may be related to or linked to the first failure. That is, for example, embodiments can generate information that can be used to identify linked or related failures, and then determine solutions to these failures before they occur in the field.

[0018] The proposed concept recognizes that slightly changing the contributing factors of a fault may lead to one or more different but linked or similar faults. By predicting a different second set of contributing factors that will lead to the occurrence of the same fault from a first set of contributing factors, variants of the fault (or similar faults) can be identified to assist in fault diagnosis and resolution. It is also proposed that the correlation between contributing factors and faults can be learned (e.g., from historical data) and thus used to predict alternative sets of contributing factors using machine learning.

[0019] Thus, various embodiments can simplify the work of exploring and resolving potential faults of medical devices by assisting in exploring linked or related faults in a guided manner.Such embodiments can provide the advantages of reduced search / exploration time and improved (ie, reduced) fault resolution time.

[0020] In other words, the embodiments propose to use machine learning to generate groups of contributing factors for each fault of medical imaging equipment. Such groups of contributing factors can be used to identify new / unforeseen faults that are linked to and / or similar to existing / known faults, thereby facilitating early or preemptive identification of one or more solutions. Therefore, the predictions provided by the embodiments can assist in fault analysis and / or diagnosis before the field deployment of the medical device. Therefore, the embodiments can be used in connection with fault resolution to support service and / or development engineers in diagnosing and resolving faults. Therefore, the proposed concept can provide improved (e.g., faster) fault resolution in medical imaging equipment.

[0021] Various embodiments may further include: a generative neural network comprising a generator and a discriminator, the generator being trained to generate a candidate set of contributing factors that may cause the occurrence of the first fault, the discriminator being trained to analyze the candidate set of contributing factors and distinguish between contributing factors that do cause the occurrence of the first fault and contributing factors that do not cause the occurrence of the first fault, and wherein the discriminator is configured to output the candidate set of contributing factors that do cause the occurrence of the first fault as the second set of contributing factors, and wherein the method further comprises:

[0022] analyzing the candidate group of contributing factors determined by the discriminator as not causing the occurrence of the first fault to determine whether the candidate group of contributing factors causes the occurrence of the first fault; and

[0023] In response to determining that the candidate set of contributing factors did not cause occurrence of the first fault, a predicted fault of the medical device caused by the candidate set of contributing factors is determined and output, wherein the predicted fault is different from the first fault.

[0024] As an example, the analyzing step may include: determining a similarity metric between the first set of contributing factor attributes and the candidate set of contributing factors; and determining whether the candidate set of contributing factors caused the occurrence of the first fault based on the determined similarity metric. For example, the similarity metric may be a cosine similarity metric, a Euclidean distance, or a Jaccard similarity coefficient.

[0025] In another example, the analyzing step may include: representing the first group of contributing factors, the candidate group of contributing factors, and the first fault as a connectivity graph; comparing the connectivity graphs; and determining whether the candidate group of contributing factors causes the occurrence of the first fault based on the comparison result.

[0026] In some embodiments, the training data for the first fault includes a description of the first fault and a corresponding first set of contributing factors, and does not include a second set of contributing factors.

[0027] In some embodiments, each set of contributing factors comprises a sequence of events.

[0028] Some embodiments may also include determining a solution to the predicted failure of the medical device; and outputting the solution.

[0029] In some embodiments, obtaining the first set of contributing factors may include: obtaining problem data describing a failure of the medical device; and analyzing the problem data to determine the first set of contributing factors that led to the failure of the medical device. In this manner, the user's actions and steps can be automatically monitored and recorded to provide data for use with machine learning algorithms. Thus, the proposed embodiments can avoid / mitigate the need for service engineers to proactively record and / or provide information about previous interactions.

[0030] The process of obtaining the first set of contributing factors can include analyzing the event history of the medical device. That is, various embodiments can utilize the act of logging / monitoring to obtain information about the various factors that led to the occurrence of the fault. Therefore, existing / conventional log file applications can be utilized to facilitate the implementation of the proposed embodiments.

[0031] In some embodiments, obtaining the first set of contributing factors may include receiving, via an input interface, a description of the contributing factors provided by a responding party in response to a request for data. The request for data may, for example, include a fault analysis questionnaire, or employ old maintenance records with comments from earlier similar faults. In this way, the responses of the responding party (e.g., an end user or a service engineer) to the questionnaire may be provided as data for identifying the first set of contributing factors. Such responses may therefore be used to identify one or more contributing factors to the first fault. Thus, various embodiments may cater for natural language descriptions of medical imaging device faults, thereby facilitating simple and intuitive data provision.

[0032] In an embodiment, the step of obtaining the first set of contributing factors may include performing a natural language processing analysis on the description of the medical device failure. By providing a natural language description of the medical imaging device failure and / or its contributing factors, various embodiments may facilitate using the provided / recorded data in a simple and intuitive manner.

[0033] In some embodiments, the machine learning algorithm may include a generative adversarial network.

[0034] Thus, one or more concepts for predicting / deriving unforeseen failures / problems from known failures are provided, and these concepts may enable preemptive identification of one or more solutions to save time, cost, and downtime of medical equipment.

[0035] According to another aspect, a computer program product is provided, wherein the computer program product comprises a computer-readable storage medium having computer-readable program code embodied therewith, the computer-readable program code being configured to perform all the steps of the proposed embodiments.

[0036] Therefore, a computer system may also be provided, comprising: a computer program product according to the proposed embodiment; and one or more processors adapted to perform the method according to the proposed concept by executing the computer-readable program code of the computer program product.

[0037] According to another aspect of the present invention, there is a system for predicting factors contributing to failure of a medical device, the system comprising one or more processors configured to:

[0038] obtaining a first set of contributing factors leading to a first failure of the medical device;

[0039] inputting a description of a first failure of the medical device and a first set of contributing factors into a machine learning algorithm, the machine learning algorithm comprising a generative adversarial network trained to predict a predicted second set of contributing factors that will lead to an occurrence of the first failure of the medical device based on the input description of the first failure and the first set of contributing factors, the second set of contributing factors being different from the first set of contributing factors, and wherein the generative adversarial network is trained to predict the second set of contributing factors using training data comprising a plurality of descriptions of the failure and, for each failure, a corresponding set of contributing factors that led to the failure;

[0040] generating a predicted second set of contributing factors using the trained machine learning algorithm in response to the input; and

[0041] Output the second set of contributing factors.

[0042] Embodiments of the system correspond to the embodiments of the method described above.

[0043] The system can be located remotely from a user device used to analyze and / or diagnose medical device failures. In this way, a user (such as a service or development engineer) can have a suitably arranged system that can receive information about contributing factors that will lead to the occurrence of a failure of the medical device at a location remotely located from the system. Thus, various embodiments can enable a user to use a local system (which can, for example, include a portable display device such as a laptop, tablet computer, mobile phone, PDA, etc.) to identify and / or explore unforeseen failures. As an example, various embodiments can provide an application for a mobile computing device, and the application can be executed and / or controlled by a user of the mobile computing device.

[0044] The system may further include: a server device including a system for predicting a failure of a medical device; and a client device including a user interface. Thus, the dedicated data processing device may be used to predict a second set of contributing factors that will lead to a first failure of the medical device, thereby reducing processing requirements or capabilities of other components or devices of the system.

[0045] The system may further comprise a client device, wherein the client device comprises a data interface, a machine learning algorithm, a processor and a display unit. In other words, a user (such as a service or development engineer) may have a suitably arranged client device (such as a laptop, tablet computer, mobile phone, PDA, etc.) that processes the received data in order to identify a predicted second set of contributing factors that will lead to the occurrence of a first failure of the medical device and generates a display control signal. By way of example only, embodiments may therefore provide a fault analysis system that enables identification of unforeseen faults that may be linked to, similar to, or related to the first fault, wherein real-time communication between the user (e.g., service engineer) and the medical device is provided and whose functionality may be extended or modified according to the proposed concept.

[0046] It will be appreciated that processing capacity may therefore be distributed throughout the system in different manners depending on predetermined constraints and / or availability of processing resources.

[0047] One or more concepts for predicting / deriving unforeseen medical device problems are provided, which can support improved medical device fault diagnosis or resolution. For example, the proposed embodiments can address the problem of having to wait for a fault to occur (e.g., after the device is deployed) before determining a solution. By identifying unforeseen problems and proactively determining one or more solutions to these problems, medical device downtime can be reduced or avoided.

[0048] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter.

[0049] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] For a better understanding of the invention, and to show more clearly how it may be put into practice, reference will now be made, by way of example, to the accompanying drawings, in which:

[0051] Figure 1 is a flow chart of a method for predicting factors contributing to failure of a medical device according to one embodiment;

[0052] Figure 2 An overview of the proposed embodiment is depicted;

[0053] Figure 3 Depicts how the parameters, their values, and event sequence tuples are combined with the noise vector as Figure 2 The conditional parameters of the model;

[0054] Figure 4 depicts a simplified block diagram of a system for predicting failure of a medical device according to one embodiment; and

[0055] Figure 5 is a simplified block diagram of a computer in which one or more components of an embodiment may be employed. DETAILED DESCRIPTION

[0056] The present invention will be described with reference to the accompanying drawings.

[0057] It should be understood that the detailed description and specific examples, while indicating exemplary embodiments of the apparatus, system, and method, are intended for illustrative purposes only and are not intended to limit the scope of the invention. These and other features, aspects, and advantages of the apparatus, system, and method of the present invention will be better understood with reference to the following description, the appended claims, and the accompanying drawings. The fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0058] Variations of the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, based on a study of the drawings, the present disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the quantifier "a" or "an" does not exclude a plurality.

[0059] It should be understood that the drawings are merely schematic and not drawn to scale.It should also be understood that the same reference numerals are used throughout the drawings to indicate the same or similar components.

[0060] The present invention proposes concepts for assisting in fault diagnosis and / or troubleshooting of medical devices. Specifically, various embodiments may provide methods and / or systems that help identify unforeseen faults from known faults, and this may facilitate the identification of one or more solutions before a fault occurs. Consequently, the proposed embodiments may reduce or avoid downtime for medical equipment.

[0061] The proposed concept can be based on the recognition that a first set of contributing factors leading to a first failure of a medical device can be used to identify a second set of contributing factors predicted to cause the first failure of the medical device. If the second set of contributing factors is different, the actual failure caused by the second set of factors may be slightly different from the first failure and / or be related to / linked to the first failure. In this way, new, previously unseen or unpredicted failures and their contributing factors can be identified, thereby facilitating preemptive action and fault resolution.

[0062] Thus, embodiments may be used in connection with medical device failure analysis and / or diagnostics, particularly during the development phase (ie, prior to deployment).

[0063] A concept for troubleshooting medical imaging equipment is also proposed, which can be used to guide engineers in resolving faults / problems.

[0064] Thus, the proposed embodiments can facilitate improved (e.g., faster and / or easier) fault analysis and / or resolution by identifying faults before they occur. Embodiments of the present invention are therefore intended to improve the search for potential faults and their solutions. Embodiments can provide efficient and predictive fault exploration mechanisms that can predict unforeseen faults. Thus, the proposed embodiments can provide a reduction in the time to diagnose and repair faults / problems.

[0065] As a further example, solutions can be prepared a priori for unforeseen / undocumented failures that may occur during the use of medical equipment. These derived failures can be considered during the development and / or deployment phase of the medical equipment. Furthermore, this can help prevent the recurrence of failures in subsequent generations of medical devices.

[0066] Figure 1 An embodiment of a computer-implemented method 100 for predicting factors contributing to a failure of a medical device is shown. Specifically, the method utilizes a first set of contributing factors that lead to a first failure of the medical device to generate (e.g., predict) a second, different set of contributing factors that are predicted to lead to the occurrence of the first failure. The second, different set of contributing factors can be used to identify one or more new failures that may be related to or linked to the first failure. In other words, the embodiment generates information that can be used to identify linked or related failures, and solutions to these failures can then be determined before they occur.

[0067] The method begins at step 110: Step 110 obtains a first set of contributing factors that led to a first failure of a medical device. Here, step 110 of obtaining the first set of contributing factors includes analyzing an event history of the medical device. For example, such an embodiment may utilize functionality of a monitoring application (e.g., activity logs / recording) to obtain information about previous events and actions.

[0068] However, in other embodiments, the step 110 of obtaining a first set of contributing factors includes: obtaining problem data describing a failure of the medical device; and analyzing the problem data to determine the first set of contributing factors that caused the failure of the medical device.

[0069] The method then proceeds to step 120, where problem data describing a first fault of the medical device is obtained. Specifically, step 120 of obtaining problem input data includes receiving, via an input interface, problem data provided by a responding party in response to a request for data. Specifically, the request for data includes a fault analysis questionnaire provided to an engineer (e.g., via a user interface). Thus, the engineer's responses to the questionnaire questions are obtained as initial problem input data and used to identify one or more characteristics of the fault. Step 120 may also include performing a natural language processing (NLP) analysis on the problem data to identify one or more characteristics of the first fault. The NLP analysis processes a natural language description of the fault, thereby facilitating identification of characteristics / aspects of the fault from information provided in a simple and intuitive manner. The description includes a high-level definition of the reported fault. For example, the fault description may include information such as an identification of a process performed by the medical device at the time of the fault, an output identification of the fault output by the medical device, or an error code output by the medical device at the time of the fault.

[0070] At step 130, the first set of contributing factors and the problem data are input into a neural network (NN)-based machine learning algorithm. The machine learning algorithm includes a generative neural network, such as a generative adversarial network, which is trained to predict a second set of contributing factors that will lead to the occurrence of a first fault of the medical device based on an input description of the first fault and the first set of contributing factors, the second set of contributing factors being different from the first set of contributing factors. The generative adversarial network is trained to predict the second set of contributing factors using training data, the training data including a plurality of fault descriptions and, for each fault, a corresponding set of contributing factors that led to the fault.

[0071] The structure of an artificial neural network (or neural network (NN) for short) is inspired by the human brain. A neural network consists of layers, each of which includes multiple neurons. Each neuron performs a mathematical operation. Specifically, each neuron can include a different weighted combination of a single type of transformation (e.g., the same type of transformation, such as sigmoid, but with different weights). As input data is processed, the mathematical operation of each neuron is performed on the input data to produce a numerical output, and the output of each layer in the neural network is sequentially fed to the next layer. The final layer provides the output.

[0072] There are several types of neural networks, such as convolutional neural networks (CNN) and recurrent neural networks (RNN). Figure 1 An exemplary embodiment of employs a generative adversarial network (GAN).

[0073] Methods for training machine learning algorithms are well known. Typically, such methods involve obtaining a training dataset comprising training input data entries and corresponding training output data entries. An initialized machine learning algorithm is applied to each input data entry to generate a predicted output data entry. The error between the predicted output data entry and the corresponding training output data entry is used to modify the machine learning algorithm. This process can be repeated until the error converges and the predicted output data entry is sufficiently similar to the training output data entry (e.g., ±1%). This is generally referred to as a supervised learning technique.

[0074] For example, the weights of the mathematical operations of each neuron can be modified until the error converges. Known methods for modifying neural networks include gradient descent, backpropagation algorithms, etc.

[0075] The training input data entries for the machine learning algorithm in method 100 correspond to example problem data and contributing factors. The training output data entries correspond to alternative contributing factors that lead to the occurrence of a failure of the medical device. That is, the machine learning algorithm is trained using a training algorithm configured to receive an array of training inputs and corresponding known outputs, wherein the training inputs include: problem data describing a known failure of the medical device and an associated set of contributing factors that lead to the occurrence of the known failure of the medical device; and wherein the corresponding known outputs include a different set of contributing factors that lead to the occurrence of the known failure of the medical device. In this manner, the machine learning algorithm is trained to output a second set of contributing factors that, when provided with a description of a first failure of the medical device and a first set of contributing factors that lead to the occurrence of the first failure, are predicted to lead to the occurrence of the first failure.

[0076] At step 140, a prediction result is generated by the machine learning algorithm in response to the input step 130. The prediction result includes a predicted second set of contributing factors.

[0077] In some examples, a generative neural network includes: a generator trained to generate a candidate set of contributing factors that may cause an occurrence of a first fault; and a discriminator trained to analyze the candidate set of contributing factors and distinguish between contributing factors that do cause an occurrence of the first fault and contributing factors that do not cause an occurrence of the first fault. The discriminator is configured to output the candidate set of contributing factors that do cause an occurrence of the first fault as a second set of contributing factors.

[0078] In these examples, step 150 includes analyzing the candidate group of contributing factors identified by the discriminator as not causing the occurrence of the first fault to determine whether the candidate group of contributing factors caused the occurrence of the first fault; and in response to determining that the candidate group of contributing factors did not cause the occurrence of the first fault, determining a predicted fault of the medical device caused by the candidate group of contributing factors and outputting the predicted fault, wherein the predicted fault is different from the first fault. Here, step 150 can adopt one or a series of alternative processes. As an example, Figure 1 The flowchart depicts two analytical processes that can be adopted by this method.

[0079] The first exemplary analysis process includes steps 152 and 154. Step 152 includes determining a value of a (cosine) similarity metric between the first set of contributing factor attributes and the candidate set of contributing factors. Step 154 includes determining whether the second set of contributing factors caused the occurrence of the first fault based on the determined value of the (cosine) similarity metric. Step 154 may, for example, involve determining a (cosine) similarity between the values of contributing factors, such as values of log file parameters.

[0080] The second exemplary analysis process includes steps 156, 157, and 158. Step 156 includes representing the first and candidate groups of contributing factors and the first fault as a connectivity graph, and comparing the connectivity graphs in step 157. Step 158 then includes determining whether the candidate group of contributing factors caused the occurrence of the first fault based on the comparison results.

[0081] Finally, in response to determining that the candidate set of contributing factors did not cause the occurrence of the first fault, the method proceeds to step 160 of determining a predicted fault of the medical device caused by a second set of contributing factors, wherein the predicted fault is different from the first fault.

[0082] like Figure 1 As shown in the dashed box in the flowchart of , the method may further include a step 170 of determining a solution to the predicted fault of the medical device. The solution may be determined in various ways. For example, the solution may be determined by consulting a database storing a corresponding solution for each of a plurality of faults. Alternatively, the solution may be determined by inputting a description of the fault into a neural network, and wherein the neural network is trained to predict a solution using training data that includes a plurality of known faults and a solution for each fault. The solution to such a fault may be determined by, for example, a service engineer consulting a service manual and determining a way to resolve the fault. The method may further include a step ( Figure 1 not shown).

[0083] Figure 1Variations of the above embodiment may include employing different methods to obtain the first set of contributing factors. For example, in an alternative embodiment, obtaining the first set of contributing factors includes receiving, via an input interface, a description of contributing factors provided by a responding party in response to a request for data. The request for data may include, for example, a fault analysis questionnaire. In yet another embodiment, obtaining the first set of contributing factors includes performing natural language processing analysis on a description of a medical device fault.

[0084] According to the proposed concept, the main elements of the exemplary embodiment can be summarized as follows:

[0085] (i) for a known (first) problem with a medical device, identifying associated parameters and events (i.e., contributing factors) during the development phase of the medical device;

[0086] (ii) Using these parameters, combinations of different values of the parameters, and sequences of events (ie, contributing factors) to identify new factors.

[0087] (iii) in step ii), using a technique based on conditional generative networks (DC-GAN) to generate new (second) factors. These may include combinations of parameters that lead to the first fault or similar unseen faults but with different values, and / or combinations of events / event sequences that lead to the same first fault or similar unseen faults;

[0088] (iv) Identify solutions to newly discovered problems

[0089] (v) Identify new paths that lead to problems or discover new, similar, and unseen problems, and use them during deployment, where the associated parameters are recorded in the actual problem;

[0090] (vi) Feed the new paths of failure / problem (ie, contributing factors) as needed to the development phase in the next iteration.

[0091] As described above, it is proposed to predict / derive unseen / new faults from a known fault set for medical equipment based on associated parameters, event sequences, etc. It is also proposed to prepare solutions for such unseen / new faults, e.g. during the development phase, in order to provide ex ante assistance to service engineers.

[0092] As another example, Figure 2 The overview of the proposed embodiment is described. The steps of the proposed embodiment are as follows:

[0093] (i) Detection of parameters, event sequences, and conditions (i.e., contributing factors) for known problems:

[0094] The first step is to understand the known / seen faults recorded for the medical device. This can include identifying the underlying parameters, their values (combinations), and the sequence of events that create the ideal conditions for the fault to occur. NLP techniques can be used to analyze such known, recorded faults 210 and identify the factors 220 that led to the fault (e.g., parameters, ranges of values, sequence of events, etc.).

[0095] The problem description 210 is pre-processed and analyzed by a set of NLP techniques (such as tokenization, stemming, POS tagging, NER), and finally a named entity set is generated to classify words and phrases based on the problem description text. Next, the system knowledge base is queried to identify parameters 220 with their value ranges, events, and conditions (causing failures).

[0096] (ii) Conditional Generative Networks for Unseen Problems:

[0097] Once the contributing factors (e.g., parameters, value ranges, event sequences, etc.) are extracted, the next step is to generate new and unseen questions. This can be achieved through a generative model, where known problem-specific parameters, their values, event sequences, and conditions can be used as conditional inputs to the model, and the model can generate a set of unseen questions 240.

[0098] The conditional deep convolutional generative adversarial network (DCGAN) model 230 is trained to achieve the following situation: where known question-specific parameters, their values, and event sequences will be the conditional parameters that guide the generation of new, unseen questions and integrate domain knowledge. As an example, Figure 3 It depicts how parameters, their values, event sequence tuples are combined with noise vectors as conditional parameters of the model, and domain-specific knowledge is used to create a new set of problems.

[0099] (iii) Verification of the generated question set:

[0100] It is important to validate the generated problem set to decide 250 whether it is a new, unseen problem, or a new path leading to a known problem. The generated (GAN model) parameters, their values, and the combination of generated event sequences can be evaluated 250 with respect to the known problem event sequences and parameter combinations (already derived in step (i)). Based on the evaluation results, a decision can be made as to whether the generated fault is a new, unseen fault, or the same fault (i.e., the known first fault) with a different path (i.e., event sequence). Similarity measurement techniques (e.g., cosine similarity) can be used for this evaluation purpose.

[0101] (iv) Prepare solutions to unseen problems 260:

[0102] Once a new set of unseen problems is identified, solutions to these problems can be pre-designed so that service engineers have knowledge of these problems and have solutions ready for them. These derived new paths to unseen or known problems can be considered (where possible) during the development / deployment phase of the medical equipment.

[0103] (v) Feed new problems and solutions into the next development cycle:

[0104] New faults designed into the existing system and their corresponding solutions are fed 270 into the next development cycle, where the next generation of development occurs. This enables faults to be removed by proactively addressing them during the design / development phase. This makes the system more robust as new generations of equipment are produced.

[0105] As an overview of another proposed implementation, Figure 4 Depicted is a simplified block diagram of a system 600 for predicting failure of a medical device.

[0106] The system includes a data interface 610 configured to obtain a first set 615 of contributing factors that lead to a first failure of a medical device. Specifically, the data interface 610 includes an analysis component configured to obtain problem data describing the first failure of the medical device and analyze the problem data to determine the first set of contributing factors that lead to the failure of the medical device via a first event path.

[0107] The machine learning algorithm 620 of the system 600 is configured to receive a description of a first fault of a medical device and a first set of contributing factor problem data and log file browsing data. As described above, the description of the first fault may include a textual description of the fault provided by a user. The machine learning algorithm 630 is trained to output a second set of contributing factors that will lead to the occurrence of the first fault of the medical device.

[0108] System 600 also includes a processor 630 configured to analyze the second set of contributing factors and the first fault of the medical device to determine whether the second set of contributing factors contributed to the occurrence of the first fault. Processor 630 is further configured to, in response to determining that the second set of contributing factors did not contribute to the occurrence of the first fault, determine a predicted fault of the medical device caused by the second set of contributing factors.

[0109] The second set of contributing factors and a description 640 of the predicted failure are provided as output 640 from the system 600 and thus can be used to predict and analyze unforeseen failures of the medical device.

[0110] Figure 5An example of a computer 1000 in which one or more components of an embodiment can be employed is shown. The various operations discussed above can utilize the capabilities of the computer 1000. For example, one or more components of a system for predicting failure of a medical device can be incorporated into any element, module, application, and / or component discussed herein. In this regard, it should be understood that the system functional blocks can be executed on a single computer or can be distributed across several computers and locations (e.g., connected via the Internet).

[0111] Computer 1000 includes but is not limited to PC, workstation, laptop computer, PDA, handheld device, server, memory etc. Generally, in terms of hardware architecture, computer 1000 may include one or more processors 1010, memory 1020 and one or more I / O devices 1030 communicatively coupled via a local interface (not shown). As known in the art, the local interface may be, for example, but not limited to, one or more buses or other wired or wireless connections. The local interface may have additional elements such as a controller, a buffer (cache), a driver, a repeater and a receiver to enable communication. In addition, the local interface may include address, control and / or data connections to enable appropriate communication between the above components.

[0112] Processor 1010 is a hardware device for executing software that may be stored in memory 1020. Processor 1010 may be virtually any custom or commercially available processor, central processing unit (CPU), digital signal processor (DSP), or auxiliary processor among several processors associated with computer 1000, and may be a semiconductor-based microprocessor (in the form of a microchip) or a microprocessor.

[0113] The memory 1020 may include any one or a combination of volatile memory elements (e.g., random access memory (RAM), such as dynamic random access memory (DRAM), static random access memory (SRAM), etc.) and non-volatile memory elements (e.g., ROM, erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic tape, compact disk read-only memory (CD-ROM), magnetic disk, floppy disk, cassette, cartridge, etc.). In addition, the memory 1020 may contain electrical, magnetic, optical, and / or other types of storage media. Note that the memory 1020 may have a distributed architecture in which various components are remotely located from one another but can be accessed by the processor 1010.

[0114] The software in the memory 1020 may include one or more separate programs, each of which includes an ordered list of executable instructions for implementing logical functions. According to an exemplary embodiment, the software in the memory 1020 includes a suitable operating system (O / S) 1050, a compiler 1060, source code 1070, and one or more applications 1080. As shown, the applications 1080 include many functional components for implementing the features and operations of the exemplary embodiments. According to an exemplary embodiment, the applications 1080 of the computer 1000 may represent various applications, computing units, logic, functional units, processes, operations, virtual entities, and / or modules, but the applications 1080 do not represent limitations.

[0115] The operating system 1050 controls the execution of other computer programs and provides scheduling, input-output control, file and data management, memory management, communication control and related services. The inventors contemplate that the application 1080 used to implement the exemplary embodiments can be applied to all commercially available operating systems.

[0116] Application 1080 may be a source program, an executable program (object code), a script, or any other entity comprising a set of instructions to be executed. When a source program, the program is typically converted via a compiler (such as compiler 1060), an assembler, an interpreter, etc., which may or may not be included in memory 1020, in order to operate properly in relation to O / S 1050. In addition, application 1080 may be written in an object-oriented programming language having classes of data and methods, or a procedural programming language having routines, subroutines, and / or functions, such as, but not limited to, C, C++, C#, Pascal, BASIC, API calls, HTML, XHTML, XML, ASP scripts, JavaScript, FORTRAN, COBOL, Perl, Java, ADA, .NET, etc.

[0117] I / O devices 1030 may include input devices such as, but not limited to, a mouse, keyboard, scanner, microphone, camera, etc. Furthermore, I / O devices 1030 may also include output devices such as, but not limited to, a printer, a display, etc. Finally, I / O devices 1030 may also include devices for communicating with input and output devices, such as, but not limited to, a NIC or modulator / demodulator (for accessing remote devices, other files, devices, systems, or networks), a radio frequency (RF) or other transceiver, a telephone interface, a bridge, a router, etc. I / O devices 1030 also include components for communicating over various networks such as the Internet or an intranet.

[0118] If the computer 1000 is a PC, workstation, intelligent device, or the like, the software in the memory 1020 may also include a basic input / output system (BIOS) (omitted for simplicity). The BIOS is a set of basic software routines that initializes and tests the hardware at startup, starts the operating system 1050, and supports data transfer between hardware devices. The BIOS is stored in some type of read-only memory, such as ROM, PROM, EPROM, EEPROM, etc., so that it can be executed when the computer 800 is activated.

[0119] When the computer 1000 is in operation, the processor 1010 is configured to execute software stored in the memory 1020, to transfer data to and from the memory 1020, and to generally control the operation of the computer 1000 according to the software. The applications 1080 and the O / S 1050 are read in whole or in part by the processor 1010, perhaps buffered within the processor 1010, and then executed.

[0120] When the application 1080 is implemented in software, it should be noted that the application 1080 can be stored on substantially any computer-readable medium for use by or in connection with any computer-related system or method. In the context of this document, a computer-readable medium can be an electronic, magnetic, optical, or other physical device or means that can contain or store a computer program for use by or in connection with a computer-related system or method.

[0121] Application 1080 can be implemented in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, a system containing a processor, or other system that can retrieve instructions from the instruction execution system, apparatus, or device and execute them. In the context of this document, a "computer-readable medium" can be any means that can store, communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. A computer-readable medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, device, or propagation medium.

[0122] Figure 1-3 Methods and Figure 4-5The system can be implemented in hardware or software or a mixture of the two (for example, as firmware running on a hardware device). To the extent that the embodiments are partially or entirely implemented in software, the functional steps shown in the process flow diagram can be performed by a suitably programmed physical computing device (such as one or more central processing units (CPUs) or graphics processing units (GPUs)). Each process shown in the flow diagram - and its individual component steps - can be performed by the same or different computing devices. According to an embodiment, a computer-readable storage medium stores a computer program including computer program code, and the computer program code is configured to cause one or more physical computing devices to perform the encoding or decoding method as described above when the program is run on the one or more physical computing devices.

[0123] Storage media may include volatile and non-volatile computer memory, such as RAM, PROM, EPROM and EEPROM, optical disks (e.g., CD, DVD, BD), magnetic storage media (e.g., hard disk and magnetic tape). Various storage media may be fixed within a computing device or may be portable so that one or more programs stored thereon can be loaded into a processor.

[0124] To the extent that the embodiments are implemented partially or fully in hardware, Figure 4-5 The blocks shown in the block diagrams may be separate physical components, or logical subdivisions of a single physical component, or may all be implemented in an integrated manner in one physical component. The functionality of a block shown in the drawings may be divided between multiple components in implementation, or the functionality of multiple blocks shown in the drawings may be combined in a single component in implementation. Hardware components suitable for use in embodiments of the present invention include, but are not limited to, conventional microprocessors, application specific integrated circuits (ASICs), and field programmable gate arrays (FPGAs). One or more blocks may be implemented as a combination of dedicated hardware that performs some functions and one or more programmed microprocessors and associated circuits that perform other functions.

[0125] Listed below are various examples of the present disclosure:

[0126] Example 1: A method for predicting failure of a medical device, the method comprising:

[0127] obtaining a first set of contributing factors leading to a first failure of the medical device;

[0128] inputting the description of the first failure of the medical device and the first set of contributing factors into a machine learning algorithm, the machine learning algorithm being trained to predict a predicted second set of contributing factors that will lead to occurrence of the first failure of the medical device, the second set of contributing factors being different from the first set of contributing factors; and

[0129] The second group of contributing factors is output.

[0130] Example 2: The method according to Example 1, further comprising:

[0131] analyzing the second set of contributing factors and the first fault of the medical device to determine whether the second set of contributing factors caused the occurrence of the first fault; and

[0132] In response to determining that the second set of contributing factors did not contribute to the occurrence of the first failure, a predicted failure of the medical device caused by the second set of contributing factors is determined, wherein the predicted failure is different from the first failure.

[0133] Example 3: The method of Example 2, wherein the analyzing comprises:

[0134] Determining a value of a similarity measure between the first set of contributing factor attributes and the second set of contributing factors; and determining whether the second set of contributing factors caused the occurrence of the first fault based on the determined value of the similarity measure.

[0135] Example 4: The method of Example 2, wherein the analyzing comprises:

[0136] representing the first set of contributing factors, the second set of contributing factors, and the first fault as a connectivity graph;

[0137] comparing the connectivity graphs; and

[0138] Based on the comparison result, it is determined whether the second group of contributing factors causes the occurrence of the first fault.

[0139] Example 5: The method according to any one of Examples 2 to 4, further comprising:

[0140] A solution to the predicted failure of the medical device is determined.

[0141] Example 6: A method according to any one of Examples 1 to 5, wherein the machine learning algorithm is trained using a training algorithm configured to receive a training input array and a corresponding known output, wherein the training input includes: problem data describing a known fault of the medical device and an associated set of contributing factors that led to the occurrence of the known fault of the medical device; and wherein the corresponding known output includes a different set of contributing factors that led to the occurrence of the known fault of the medical device.

[0142] Example 7: The method of any one of Examples 1 to 6, wherein obtaining the first set of contributing factors comprises:

[0143] obtaining problem data describing the failure of the medical device; and

[0144] The problem data is analyzed to determine a first set of contributing factors to the failure of the medical device.

[0145] Example 8: The method of any one of Examples 1 to 7, wherein obtaining the first set of contributing factors comprises:

[0146] An event history of the medical device is analyzed.

[0147] Example 9: The method of any one of Examples 1 to 8, wherein obtaining the first set of contributing factors comprises:

[0148] receiving, via an input interface, a description of contributing factors provided by a responding party in response to a request for data,

[0149] And optionally, the request for data includes a fault analysis questionnaire.

[0150] Example 10: The method of any one of Examples 1 to 9, wherein obtaining the first set of contributing factors comprises:

[0151] A natural language processing analysis is performed on the description of the malfunction of the medical device.

[0152] Example 11: A method according to any one of Examples 1 to 10, wherein the machine learning algorithm includes a generative adversarial network.

[0153] Example 12: A computer program comprising code means for implementing the method according to any one of Examples 1 to 11 when the program is run on a processing system.

[0154] Example 13: A system for predicting failure of a medical device, the system comprising:

[0155] a data interface configured to obtain a first set of contributing factors that led to a first failure of the medical device; and a machine learning algorithm configured to receive a description of the first failure of the medical device and the first set of contributing factors, the machine learning algorithm being trained to output a predicted second set of contributing factors that will lead to an occurrence of the first failure of the medical device.

[0156] Example 14: The system of Example 13, further comprising:

[0157] a processor device configured to analyze the second set of contributing factors and the first fault of the medical device to determine whether the second set of contributing factors caused the occurrence of the first fault; and in response to determining that the second set of contributing factors did not cause the occurrence of the first fault, determine a predicted fault of the medical device caused by the second set of contributing factors, wherein the predicted fault is different from the first fault.

[0158] Example 15: The system of Example 10, wherein the data interface includes an analysis component configured to obtain problem data describing a failure of the medical device and analyze the problem data to determine a first set of contributing factors that caused the failure of the medical device via a first event path.

[0159] Based on a study of the drawings, the present disclosure and the appended claims, those skilled in the art can understand and implement variations of the disclosed embodiments when practicing the claimed invention. In the claims, the word "comprising" does not exclude other elements or steps, and the quantifier "a" or "an" does not exclude a plurality. A single processor or other unit can implement the functions of several items recited in the claims. The fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to profit. If the computer program is as described above, it can be stored / distributed on a suitable medium, such as an optical storage medium or solid-state medium supplied together with other hardware or as part of other hardware, but can also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems. If the word "suitable for" is used in a claim or specification, it should be noted that the word "suitable for" is intended to be equivalent to the word "configured to". Any figure marks in the claims should not be interpreted as limiting the scope.

[0160] The flow charts and block diagrams in the accompanying drawings illustrate the architecture, functions and operations of the possible implementations of the systems, methods and computer program products according to various embodiments of the present invention. In this regard, each frame in the flow chart or block diagram can represent a module, segment or portion of an instruction, which includes one or more executable instructions for realizing a specified logical function. In some alternative implementations, the functions mentioned in the frame can occur in a sequence that is separated from the sequence indicated in the figure. For example, the two frames shown in succession can actually be performed substantially simultaneously, or the frames can sometimes be performed in an opposite order, depending on the functions involved. It will also be noted that each frame of the block diagram and / or flow chart illustration and the combination of the frames in the block diagram and / or flow chart illustration can be implemented by a dedicated hardware-based system that performs a specified function or action or performs a combination of dedicated hardware and computer instructions.

Claims

1. A method for predicting factors contributing to failure of a medical device, the method comprising: obtaining a first set of contributing factors leading to a first failure of the medical device; inputting the description of the first fault of the medical device and the first set of contributing factors into a machine learning algorithm, the machine learning algorithm comprising a generative neural network trained to predict a predicted second set of contributing factors that will lead to occurrence of the first fault of the medical device based on the input description of the first fault and the first set of contributing factors, the second set of contributing factors being different from the first set of contributing factors, and wherein the generative neural network is trained to predict the second set of contributing factors using training data comprising a plurality of fault descriptions and, for each fault, a corresponding set of contributing factors that led to the fault; generating a second set of contributing factors to the prediction using a trained machine learning algorithm in response to the input; and The second group of contributing factors is output.

2. The method according to claim 1, wherein The generative neural network includes a generative adversarial network.

3. The method according to claim 1 or claim 2, wherein: The generative neural network includes a generator and a discriminator, the generator being trained to generate a candidate group of contributing factors that may cause the occurrence of the first fault, and the discriminator being trained to analyze the candidate group of contributing factors and distinguish between contributing factors that do cause the occurrence of the first fault and contributing factors that do not cause the occurrence of the first fault, and wherein the discriminator is configured to output the candidate group of contributing factors that do cause the occurrence of the first fault as the second group of contributing factors, and wherein the method further comprises: analyzing a candidate group of contributing factors determined by the discriminator as not causing the occurrence of the first fault to determine whether the candidate group of contributing factors causes the occurrence of the first fault; and In response to determining that the candidate set of contributing factors did not cause occurrence of the first fault, a predicted fault of the medical device caused by the candidate set of contributing factors is determined and output, wherein the predicted fault is different from the first fault.

4. The method according to claim 3, wherein: The analysis includes: determining a similarity measure between the first set of contributing factor attributes and the candidate set of contributing factors; and Based on determining the similarity metric, it is determined whether the candidate group of contributing factors causes the occurrence of the first fault.

5. The method according to claim 2, wherein: The analysis includes: representing the first group of contributing factors, the candidate group of contributing factors, and the first fault as a connectivity graph; comparing the connectivity graphs; and Based on the comparison result, it is determined whether the candidate group of contributing factors causes the occurrence of the first fault.

6. The method according to claim 1, wherein The training data for the first fault includes a description of the first fault and the corresponding first set of contributing factors, and does not include the second set of contributing factors.

7. The method according to claim 1, wherein Each set of contributing factors consists of a sequence of events.

8. The method according to any one of claims 3 to 7, further comprising: determining a solution to the predicted failure of the medical device; and Output the solution.

9. The method according to any one of claims 1 to 8, wherein Obtaining the first set of contributing factors includes: obtaining problem data describing the failure of the medical device; and The problem data is analyzed to determine a first set of contributing factors to the failure of the medical device.

10. The method according to any one of claims 1 to 8, wherein Obtaining the first set of contributing factors includes: An event history of the medical device is analyzed.

11. The method according to any one of claims 1 to 10, wherein Obtaining the first set of contributing factors includes: receiving, via an input interface, a contribution factor description provided by a responding party in response to a request for data, And optionally, the request for data includes a fault analysis questionnaire.

12. The method according to any one of claims 1 to 11, wherein Obtaining the first set of contributing factors includes: A natural language processing analysis is performed on the description of the malfunction of the medical device.

13. A computer program comprising code modules for implementing the method according to any one of claims 1 to 12 when the program is run on a processing system.

14. A system for predicting factors contributing to failure of a medical device, the system comprising one or more processors configured to: obtaining a first set of contributing factors leading to a first failure of the medical device; The description of the first fault of the medical device and the first set of contributing factors are input into a machine learning algorithm, the machine learning algorithm comprising a generative neural network trained to predict a predicted second set of contributing factors that will lead to the occurrence of the first fault of the medical device based on the input description of the first fault and the first set of contributing factors, the second set of contributing factors being different from the first set of contributing factors, and wherein The generative neural network is trained to predict the second set of contributing factors using training data, the training data comprising a plurality of fault descriptions and, for each fault, a corresponding set of contributing factors that caused the fault; generating a second set of contributing factors to the prediction using a trained machine learning algorithm in response to the input; and The second group of contributing factors is output.

Images