A self-correction approach to annotations of data tools using feedback mechanisms

Abstract

In a monitoring method for generating a maintenance alert, a component ID identifying a medical imaging device component in an electronic medical imaging device manual (25, 26, 27, 28) is extracted. Operating parameters and associated operating parameter ranges for the medical imaging device component are also extracted from the manual based on numerical values, parameter items identifying operating parameters, and link items or symbols connecting the numerical values ​​and parameter items indicating equality or inequality. The operating parameter ranges are formulated as decision rules (36) that are applied to log data (40) generated by the monitored medical imaging device (2) to detect out-of-range log data generated by the monitored medical imaging device. A maintenance alert (24) is displayed on a display (18) in response to the detected out-of-range log data. The maintenance alert is generated based on the out-of-range log data and is associated with the component ID contained in the out-of-range log data.

Description

Technical Field

[0001] The following generally relates to medical imaging equipment maintenance techniques, equipment monitoring techniques, predictive maintenance techniques, and related techniques. Background Art

[0002] Medical devices used in the healthcare industry (e.g., Magnetic Resonance Imaging (MRI) scanners, Computed Tomography (CT) scanners, Positron Emission Tomography (PET) scanners, gamma cameras used in Single Photon Emission Computed Tomography (SPECT), Image Guided Therapy (iGT) systems, and other medical imaging devices or electrocardiographs (ECG) or patient monitor devices, etc.) should be in good working condition, thereby ensuring that doctors and patients receive the right information for medical diagnosis and patient monitoring, etc. Medical imaging devices are expensive to replace and play a vital role in diagnosis. Any downtime of these devices results in a loss of revenue for the medical institution, a decrease in the quality of patient treatment, and delays in patient treatment. Therefore, a focus is placed on minimizing the downtime of medical imaging devices and ensuring uninterrupted operational status while maintaining performance quality.

[0003] Predictive maintenance is an important component of minimizing downtime of medical imaging equipment. The approach entails predicting and proactively repairing or otherwise remedying the likely failure of medical imaging equipment components based on machine logs and usage history. In this way, maintenance can be proactively performed to minimize or eliminate downtime and the impact on patient care.

[0004] However, predictive maintenance is difficult to implement in practice. Modern medical imaging equipment generates a large amount of log data, which is on the order of GB or TB or more. In addition, due to the time lag between the log data and the subsequent component failure, it is difficult to proactively identify log data that statistically indicates a possible component failure. Moreover, it is difficult to determine which log data can be used to predict component failure. Conventionally, prior information is used to identify this type of bad log data that indicates a possible component failure. Typically, the prior information is provided by subject matter experts who have specialized technical knowledge of medical imaging equipment components, their performance ranges, and possible problems. Generating a knowledge engine for performing predictive maintenance is typically a laborious process that requires input from subject matter experts with extensive knowledge of medical imaging equipment and its components to develop decision rules that can be applied to the log data of the monitored medical imaging equipment.

[0005] Another problem is that the resulting predictive maintenance system is static. This is a problem because medical imaging equipment manufacturers are constantly adding new products and improving existing product lines. These changes will not be reflected in the knowledge engine used for predictive maintenance. Therefore, manual updates to the knowledge engine must be performed frequently, which again requires extensive input from subject matter experts.

[0006] Certain improvements are disclosed below. Summary of the invention

[0007] In some embodiments disclosed herein, a non-transitory storage medium stores instructions that can be read and executed by an electronic processor to perform a monitoring method for generating a maintenance alert. The monitoring method includes: extracting a component identification (component ID) identifying a medical imaging device component and an operating parameter of the identified medical imaging device component and an associated operating parameter range from an electronic medical imaging device manual; generating a knowledge engine by an operation including formulating the operating parameter range as a set of decision rules for classifying medical imaging device log data; and applying the knowledge engine to the log data generated by the monitored medical imaging device to detect out-of-range log data generated by the monitored medical imaging device and generate a maintenance alert in response to the detected out-of-range log data, wherein the maintenance alert is associated with the component ID contained in the detected out-of-range log data.

[0008] In some embodiments disclosed herein, a monitoring device for generating a maintenance alert is disclosed. The monitoring device includes: a display; a non-transient storage medium as described in the previous paragraph; and an electronic processor operatively connected to the display and the non-transient storage medium to perform the monitoring method, the monitoring method also including displaying the generated maintenance alert on the display.

[0009] In some embodiments disclosed herein, a monitoring method for generating a maintenance alarm performed by an electronic processor is disclosed. A component identification (component ID) identifying a medical imaging device component in an electronic medical imaging device manual is extracted. The operating parameters and associated operating parameter ranges of the identified medical imaging device component are extracted from the electronic medical imaging device manual based on the numerical values, parameter items, and link items or symbols connecting the numerical values ​​and the parameter items indicating equality or inequality of the operating parameters. The operating parameter range is formulated as a decision rule for classifying medical imaging device log data based on whether the value of the associated operating parameter is outside the operating parameter range. The decision rule is applied to the log data generated by the monitored medical imaging device to detect out-of-range log data generated by the monitored medical imaging device, and a maintenance alarm is displayed on a display in response to the detected out-of-range log data. The maintenance alarm is generated based on the out-of-range log data and is associated with the component ID contained in the out-of-range log data.

[0010] In some embodiments disclosed herein, a monitoring device for generating a maintenance alert is disclosed. The monitoring device includes: a display; an electronic processor operatively connected to the display; and a non-transitory storage medium storing instructions that can be read and executed by the electronic processor to perform a monitoring method for generating a maintenance alert. In the monitoring method, a decision rule is generated, the decision rule being used to classify medical imaging device log data based on whether a parameter value of an operating parameter of a medical imaging device component identified by a component identification (component ID) in the log data is within an operating parameter range for the corresponding operating parameter. The operating parameter range for the corresponding operating parameter is extracted from an electronic medical imaging device manual. The decision rule is applied to the log data generated by the monitored medical imaging device to detect out-of-range log data generated by the monitored medical imaging device. In response to the detected out-of-range log data, the display is controlled to present a maintenance alert, wherein the maintenance alert is associated with the component ID contained in the detected out-of-range log data.

[0011] One advantage resides in providing a monitoring device for generating maintenance alerts for one or more monitored medical imaging devices, wherein a knowledge engine of the monitoring device is automatically generated.

[0012] Another advantage resides in providing a monitoring device wherein a knowledge base is continuously and efficiently updated based on a maintenance log of a monitored medical imaging device.

[0013] Another advantage resides in providing computationally efficient knowledge base generation and updating for such monitoring devices.

[0014] Another advantage resides in providing a monitoring device in which a knowledge base can be developed without input from subject matter experts.

[0015] Other benefits include reducing the overall time to build predictive models, reducing human error by removing subjectivity from the labeling process, and automatic feedback to relearn and self-correct decision rules used to generate maintenance alerts.

[0016] A given embodiment may provide none, one, two, more, or all of the aforementioned advantages, and / or may provide other advantages as will become apparent to one of ordinary skill in the art upon reading and understanding the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The invention may take form in various components and arrangements of components, and in various steps and arrangements of steps.The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting the invention.

[0018] Figure 1 A monitoring device for performing predictive maintenance including generating maintenance alerts for a monitored medical imaging device based on log data generated by the monitored medical imaging device is diagrammatically illustrated, wherein decision rules for generating the maintenance alerts are automatically generated.

[0019] Figure 2 Diagrammatically illustrates a portion of a medical imaging device manual.

[0020] Figure 3 Diagrammatically illustrates the Figure 2 Table of components and key figures exported from the manual section shown.

[0021] Figure 4 , Figure 5 and Figure 6 Diagrammatically illustrates aspects of generating decision rules based on the medical imaging device manual content described herein.

[0022] Figure 7 and Figure 8 Diagrammatically illustrates aspects of mapping medical imaging device log data to decision rules of the knowledge engine described herein.

[0023] Fig. 9 Diagrammatically illustrates the Figure 1 The monitoring device is appropriately executed to construct a method for decision rules of a knowledge engine.

[0024] Fig.10 and Fig.11 Diagrammatically illustrates Fig. 9Aspects of the reader module components of the method. DETAILED DESCRIPTION

[0025] In an embodiment disclosed herein, an electronic medical imaging device manual is used to extract a component identification (component ID) identifying a medical imaging device component and an operating parameter of the identified medical imaging device component and an associated operating parameter range. The electronic medical imaging device manual can be in substantially any machine-readable form and can be an online manual, a printed manual scanned into pdf (or other electronic format) and processed by optical character recognition (OCR) to generate machine-readable text, etc. The electronic medical imaging device manual may, for example, include one or more of a maintenance manual, (one or more) manufacturer specifications, a user manual, an operating reference manual, etc. A knowledge engine is then generated by an operation including formulating an operating parameter range as a set of decision rules for classifying medical imaging device log data. For example, if the input log data indicates a value for an operating parameter that is outside an operating parameter range associated with the operating parameter, the decision rule may classify the input log data as bad (i.e., out-of-range) log data. The knowledge engine is then applied to the log data generated by the monitored medical imaging device to detect out-of-range log data generated by the monitored medical imaging device and generate a maintenance alert in response to the detected out-of-range log data. Maintenance alerts are appropriately associated with component IDs contained in the detected out-of-range log data, and in some embodiments may be formulated as maintenance advisories. For example, if the temperature of a component is indicated as being outside of the operating temperature range for that component, an alert may be formulated as "<component ID>'s operating temperature may be above its recommended operating temperature. It is recommended that the temperature of <component ID> be checked."

[0026] Another aspect is that the accuracy of the generated maintenance alert can be determined based on feedback extracted from the maintenance log of the monitored medical imaging device, and the knowledge engine can be updated by adjusting the decision rule based on the determined accuracy of the generated maintenance alert. For example, the accuracy of the generated maintenance alert can be determined based on feedback extracted from the maintenance log within a time interval extending from the time of generation of the maintenance alert to an end time that depends on (and may be equal to) the mean time to failure (MTTF) of the medical imaging device component identified by the component ID associated with the maintenance alert.

[0027] By the disclosed method, the need to develop and manage manuals to derive a set of decision rules for performing predictive maintenance is reduced or eliminated.

[0028] refer to Figure 1, a monitoring device for generating a maintenance alert for a medical imaging device 2 is disclosed. The illustrative medical imaging device 2 is a PET / CT scanner that includes a positron emission tomography (PET) gantry 4 and a transmission computed tomography (CT) gantry 6 placed side by side so that a patient can be loaded into either gantry 4, 6 via a commonly used robotic patient support or couch 8. Such a PET / CT scanner 2 is commonly used for many diagnostic and clinical tasks (e.g., oncology, cardiology, neurology, etc.) because the CT imaging performed by the CT gantry 6 can provide anatomical information that is complementary to the functional information provided by the PET imaging performed by the EPT gantry 4. As just one illustration of this synergy, the CT images can be used to generate attenuation maps that are used to perform attenuation correction during PET image reconstruction. The PET / CT scanner 2 is a highly complex medical imaging device having thousands of components, such as, as a few non-limiting examples, tens or hundreds of PET detector modules, an X-ray tube and tens or hundreds of X-ray detector modules (components of the CT gantry 6), mechanical components for rotating the X-ray tube and the X-ray detector array around the patient, numerous robotic mechanisms for the patient support or couch 8, electrical components (e.g., distribution transformers for delivering power to the PET / CT 2), etc. (note that in the Figure 1 Detailed medical imaging device components are not shown in the figure). Each component typically has one, two, three or more operating parameters monitored by appropriate sensors. For example, each PET detector module may include a temperature sensor for outputting the module operating temperature, an electrical sensor for monitoring dark current and / or other electrical operating parameters, etc. The robotic component of the couch 8 may include a position encoder for monitoring the positioning of the robotic actuator. The CT gantry may include a sensor for monitoring the rotation speed and encoding the rotational angular position. The distribution transformer may include sensors for monitoring the operating power factor, input voltage, and output voltage, etc. Moreover, these are only a few non-limiting illustrative examples. Although the illustrative monitored medical imaging device 2 is a PET / CT scanner, it will be appreciated that in more general cases, the monitored medical imaging device may be any type of medical imaging device with sufficient complexity to prove the continuous monitoring and timely generation of maintenance alarms disclosed herein. For example, the monitored medical imaging device may be a PET / CT scanner, a SPECT / CT scanner, an independent gamma camera, an independent PET scanner, an independent CT scanner, a magnetic resonance (MR) imaging scanner, a PET / MR scanner, an image-guided therapy (iGT) device, etc. Furthermore, it will be appreciated that the disclosed predictive maintenance monitoring devices and methods may be applied to monitor many such (possibly different types) of medical imaging devices in parallel.

[0029] Continue to refer Figure 1 , the illustrative monitoring device includes a server computer 10 or other electronic processor (e.g., a desktop computer with sufficient computing power) operatively connected to a non-transitory storage medium 12 that stores instructions that can be read and executed by the electronic processor 10 to perform the monitoring method disclosed herein. The illustrative monitoring device also includes a monitor interface device 14, which in the illustrative embodiment is embodied by a computer 16 acting as a medical imaging device controller 14 for controlling the monitored medical imaging device 2 (e.g., to operate a robotic support 8 to load a patient into an appropriate rack, select and run an imaging sequence or protocol, perform reconstruction on the acquired imaging data to generate an image and display the image on a display 18 of the computer, etc.). In other embodiments, the monitor interface device can be a separate computer or can be integrated with the computer 10 that performs the monitoring method. The monitor interface device includes a display 18 for presenting maintenance alerts or other information, and can optionally include one or more user input devices (e.g., an illustrative keyboard 19) for receiving user input.

[0030] The non-transitory storage medium 12 may be embodied in various ways, such as as a hard drive, a RAID array or other magnetic storage medium, a solid state drive (SSD) or other electronic storage medium, an optical disk or other optical storage medium, various combinations thereof, etc. Additionally, it will be appreciated that the illustrative electronic processors 10, 16 may also be embodied and / or combined in various other ways, and / or the various non-transitory storage media may be embodied in various ways, such as by being linked via an electronic data network, etc. For example, the electronic processor 10 may be implemented as a cloud computing resource comprising a temporary combination of multiple server computers.

[0031] The electronic processor 10 reads and runs instructions stored on the non-transitory storage medium 12 to perform a monitoring method, which includes implementing a knowledge engine builder 20 that builds and optionally then adaptively maintains or updates a knowledge engine 22 that includes a set of decision rules for generating maintenance alerts 24. To this end, the knowledge engine builder 20 utilizes input from available electronic medical imaging device operation data. Common forms of electronic medical imaging device maintenance and operation data are electronic manuals (e.g., an explanatory maintenance manual 25), a collection of explanatory manufacturer specifications 26, an explanatory user manual 27, and an explanatory operating reference manual 28. These are merely illustrative names, and it will be appreciated that various medical imaging device manufacturers and users (e.g., hospitals) may use different names for electronic medical imaging device manuals. Likewise, the manual content may be distributed in various ways among one or more such manuals, for example, in some embodiments, a single manual may cover the combined content of an explanatory user manual 27 and an operating reference manual 28, and / or the maintenance manual may include the manufacturer's specifications as one or more appendices to the maintenance manual rather than as a separate manufacturer's specification document, etc. The maintenance and operation data of the electronic medical imaging device may also include maintenance clauses compiled into electronic form (e.g., electronic maintenance logs maintained by maintenance engineers), manual updates (e.g., basic system parameter updates or operating parameter updates due to system / software upgrades, etc.), external data (e.g., online maintenance and operation data available on the website of the medical imaging device supplier), etc. The various electronic medical imaging device maintenance and operation data 25, 26, 27, 28 are preferably specific to the brand and model of the medical imaging device 2 to be monitored, but this requirement may be relaxed in the case where certain systems or subsystems are shared by medical imaging devices of different brands and / or models of a specific type.

[0032] The knowledge engine builder 20 processes the content of the electronic medical imaging equipment maintenance and operation data 25, 26, 27, 28 to extract the component identification (component ID) identifying the medical imaging equipment component, and extract the operating parameters of the identified component, and extract the operating parameter range associated with the corresponding operating parameter. For this purpose, the manual is assumed to be in electronic form with machine-readable text. (If this is not the case initially, optical scanning, photocopying, photography, etc. can be used to generate digital images of the manual pages, and then OCR is performed to convert the paper manual into an electronic format with machine-readable text). Natural language processing (NLP) 30 is performed on the text of the electronic medical imaging equipment maintenance and operation data 25, 26, 27, 28 to segment the text into individual tokens (e.g., words), delete common words that do not contain information (e.g., "the", "a", etc.), perform stemming and lemmatization operations, etc. NLP 30 can include parsing of the text and tables of the electronic medical imaging equipment manual to depict semantic units including sentences, paragraphs, list items, table rows, etc.

[0033] Component Entity Recognition (CER) 32 is applied to identify component IDs that identify medical imaging device components in the NLP processed text. CER 32 can use any type of named entity recognition (NER) algorithm or combination of NER algorithms, for example, using a vocabulary of a specific domain to identify component IDs, identifying component IDs based on factors such as part of speech (where NLP 30 includes grammatical analysis), etc. Although CER 32 is preferably a fully automatic process, it is also envisioned in alternative embodiments that it is semi-supervised, such as presenting uncertain component IDs to a user for confirmation or rejection. CER 32 also performs operating parameter extraction to extract operating parameters of the medical imaging device component identified by the component ID and associated operating parameter ranges. In one method, operating parameters are identified by identifying numerical values ​​and parameter items connected by link items or symbols indicating equality or inequality, and the parameters identified by the parameter items are associated with medical imaging device components whose component IDs appear in the same semantic units as the parameter items. For example, for the sentence "The PET detector module should be maintained at a temperature below 100°C.", it can be processed by identifying the numerical value (100°C) and the parameter term (temperature), connecting these numerical values ​​and parameter terms by a link term or symbol indicating equality or inequality (as shown below), and associating the operating parameter "temperature" with the component ID (PET detector) that appears in the same semantic unit (same sentence) as the parameter term (temperature). The parameter terms can be identified using NER to identify "named entities" representing the operating parameters, such as using domain-specific vocabulary, parts of speech (if available, in the case of grammatical analysis), etc. to identify the "named entities" representing the operating parameters. Again, this is preferably a fully automated approach, but in some embodiments, semi-supervised parameter term extraction is also envisioned, such as by presenting the uncertain parameter name to the user for confirmation or rejection. The operating parameter range associated with the operating parameter is determined based on the numerical value and the equality or inequality indicated by the link term or symbol. In the last example, the operating range of the extracted operating parameter "temperature" for the component ID "PET detector" can be expressed as "temperature < 100°C". The knowledge engine 22 is then generated by operations including formulating the operating parameter ranges as a set of decision rules 36 for classifying the medical imaging device log data as good data (e.g., within the operating range) or bad log data (e.g., outside the operating range). Advantageously, extracting the component IDs and operating parameters and associated operating parameter ranges and generating the knowledge engine 22 does not require receiving input from a subject matter expert.

[0034] The monitoring of the monitored medical imaging device 2 utilizes a machine log 40 typically generated by the medical imaging device. Figure 1In the present invention, the medical imaging device controller 14 generates a machine log based on the readings of various sensors of the components, subsystems and systems of the monitored medical imaging device 2. The machine log 40 can be stored on the non-transient storage medium 12 using a log record format, which generally adopts a standardized syntax and format for the machine type and manufacturer of the medical imaging device 2. Therefore, based on the known standard log representation, the extraction-transformation-loading (ETL) 42 of the machine log data 40 will be very simple. The component ID and the operating parameters and their log records are identified in the ETL processed log data, and the knowledge engine 22 applies the decision rules 36 to the information to identify any operating parameters whose values ​​exceed the corresponding operating parameter range. The identified out-of-range operating parameters trigger the generation of a maintenance alert 24. In other words, the knowledge engine 22 is applied to the log data generated by the monitored medical imaging device 2 to detect bad log data generated by the monitored medical imaging device 2 and generate a maintenance alert 24 in response to the detected bad log data. The maintenance alarm is associated with the component ID contained in the detected bad log data, for example, the log data will report the temperature reading associated with the component (identified by its component ID) whose reading was taken. Optionally, the maintenance alarm can be generated only after the out-of-range parameter value is detected within a certain time interval (which may depend on the sensor) to avoid excessive triggering of maintenance alarms due to occasional inaccuracies in sensor readings. It is also envisioned that the maintenance alarm 24 is coupled to the out-of-range log data (or more specifically, the out-of-range parameter value of the log data) based on a ratio other than 1:1. The maintenance alarm 24 can be generated in response to a combination of two or more parameters detected in the out-of-range log data that are outside the corresponding operating parameter ranges of the two or more parameters. That is, a decision rule can be applied to detect a combination of two or more parameters in the out-of-range log data that are outside the respective operating parameter ranges of the two or more parameters. For example, a maintenance alarm can be generated in response to the (possibly weighted) sum of different out-of-range parameters (for example, if the amount of two parameters is out of range, and these amounts individually do not trigger a maintenance alarm, but the combination of the out-of-range parameters will trigger an alarm, this is the case). An alarm may be triggered with a lower tolerance, for example, considering that parameters are generally identified as parameter A and parameter B, a maintenance alarm may be issued if: A>5%; B>10% or A>3% and B>8%. In this example, the tolerance weight of parameter A is higher than the tolerance weight of parameter B (40% vs. 20% in the combined tolerance variation).

[0035] The maintenance alert 24 is presented to the user, for example, by displaying the maintenance alert 24 in an "Alerts" window shown on the display 18 of the monitor interface device 14. For example, the maintenance alert may be presented as a maintenance advisory, such as, "It is recommended to check the PET detector temperature, the most recent reading has exceeded the recommended upper limit of 100°C". In some embodiments, the maintenance alerts may be graded with respect to severity or urgency, such as based on the extent to which the parameter value is out of range (e.g., a PET detector reading of 102°C would trigger the lowest level of alarm, while a reading of 115°C would trigger a higher level of alarm) and / or based on the criticality of the component (e.g., an out-of-range parameter reading for the PET detector would generate a higher level of alarm than an out-of-range parameter reading for a bearing of the patient support 8). In some embodiments, the processing 30, 32 of the manual may include detecting emergency symbols contained in the text of the manual itself. For example, it is not uncommon for the manual to emphasize critical parameter ranges through keywords such as "important" - when such keywords are detected in conjunction with the extracted decision rule, the decision rule will be marked as having high importance, and any maintenance alert triggered by such a rule will be assigned a high level of alarm. The maintenance alert 24 may also optionally indicate severity or urgency in other ways, such as by potential impact (e.g., performance degradation, short-term failure, long-term failure, potential clinical impact, etc.). In addition to being presented to the user, the maintenance alert 24 may also optionally be forwarded to the medical imaging device vendor or other maintenance service provider so that a service call and / or parts dispatch may be adjusted to be accurate, if possible.

[0036] In some embodiments, feedback from the log is used to adaptively adjust or tune the decision rules 36 applied by the knowledge engine 22. This is done by utilizing a maintenance log 44 that is typically maintained for the medical imaging device 2, for example, based on maintenance records manually generated by maintenance personnel and / or automatically generated by system software that, for example, detects the replacement of various components and timestamps them. The maintenance log 40 can be stored on a non-transitory storage medium 12 using a log record format that typically employs a standardized syntax and format for the machine type and manufacturer and / or a standardized syntax and format used by a maintenance organization. Therefore, based on known standard log symbols, the ETL 46 of the maintenance log data 44 is simple. Typically, the accuracy of the generated maintenance alert 24 is determined in a feedback analysis 48 based on feedback extracted from the maintenance log 44, and the knowledge engine 22 is updated by adjusting the decision rules 36 based on the determined accuracy of the generated maintenance alert. For example, the accuracy of a generated maintenance alert may be determined based on feedback extracted from the maintenance log 44 over a time interval extending from the time of generation of the maintenance alert to an end time dependent on the mean time to failure (MTTF) of the medical imaging device component identified by the component ID associated with the maintenance alert. To account for statistical differences, the end time may be a scaled value, e.g., 1.5×MTTF. If maintenance of the subject component is recorded in the maintenance log 44 during the time interval, or a failure of the subject component is recorded in the maintenance log 44, the maintenance alert may be considered accurate. On the other hand, if no maintenance is detected and the subject component does not fail within the time interval, the maintenance alert may be considered inaccurate because the maintenance personnel believe that taking action on the maintenance alert is inappropriate and the maintenance alert cannot accurately predict the impending failure of the subject component.

[0037] If the feedback analysis 48 determines that the maintenance alert is accurate, no adjustment or remediation is performed. On the other hand, if the feedback analysis 48 determines that the maintenance alert is inaccurate, this information is used to adjust or remediate the decision rule that triggered the maintenance alert. In the preferred fully automatic adjustment embodiment, the adjustment or remediation of the inaccurate decision rule may require deleting the inaccurate decision rule or adjusting the operating parameter range of the inaccurate decision rule. Such an adjustment may take the form of a percentage change, for example, if the inaccurate decision rule triggers a maintenance alert when a maximum parameter value threshold is exceeded, the threshold may be increased by 5% (or increased by 10%, or increased by some other selected adjustment increment). Similarly, if the inaccurate decision rule triggers a maintenance alert when a minimum parameter value threshold is exceeded, the threshold may be reduced by 5% (or reduced by 10%, or reduced by some other selected adjustment increment). In a variant semi-supervised embodiment, the inaccurate decision rule (possibly along with past instances that have triggered an alert) may be presented to the user for review, and the user may choose to manually adjust the inaccurate decision rule, or may choose to delete the rule entirely, or may choose to maintain the rule without any adjustment.

[0038] refer to Figure 1 , provides an overview of a monitoring device for monitoring a medical imaging device 2, and some additional embodiments and examples are presented below.

[0039] Regarding the processing operations 30, 32, there are different types of manuals associated with the medical imaging device 2, which can provide a lot of information about the device. The user manual 27 can tell the functions of the medical imaging device and how to use it, its use case boundary conditions, and this helps to ensure that the user practices and uses it correctly. The maintenance manual 25 can give more detailed information at the component level, the service life based on the usage, the absolute life, symptoms and behaviors, the actions that need to be taken for each problem, and other key parameters and their boundaries. Similarly, the manufacturer's specification 26 is an important part of cost and quality control for testing, calibration and other measurement processes. Integrating the information in the above manuals can provide comprehensive information to provide maintenance alerts, thereby facilitating predictive maintenance of medical imaging devices and their components. The data absorbed from the user manual 27, the maintenance manual 25 and the manufacturer's specification 26 are parsed to extract context-related information from the documents. Component entity recognition (CER) 32 is performed to identify the name of the component (i.e., component ID) and the name of the operating parameter in order to identify the relevant operating parameters for these components. Co-reference is performed to identify specific sentences where the association is unclear. Correlation analysis is performed to identify patterns associated with specific components that need to be mapped. Additionally, each relevant sentence associated with the component is extracted and listed.

[0040] refer to Figure 2 and Figure 3 , an example is shown. Figure 2 An example of manual content for a dedicated distribution transformer from the CT Installation Guide is shown. Figure 3 Shows from Figure 2 The component ID extracted from the content of (i.e., component ID "dedicated distribution transformer") and the key indicators extracted for this component. All other extracted component IDs are processed in a similar manner to obtain a structured table ( Figure 3 is part of this table), which is a structured table containing all components (or more specifically, component IDs) and associated key indicators that need to be monitored for the smooth operation of the medical imaging device.

[0041] Reference now Figure 4 Once the data has been extracted and the component table has been created, a composite is made for the next step that allows for data classification. Creating decision rules 36 for the knowledge engine 22 requires combining the data extracted from the different manuals in an efficient and effective manner for each monitored component of the medical imaging device. For each component, the data from the maintenance manual, manufacturer specifications, and user manual are merged into a table form from the large amount of input cleaned in operations 30, 32. This process is performed by the following operations: (1) Obtain system details from the user manual, such as system manufacturer, model, etc. Each system can be associated with a unique ID; (2) For the system details from step (1), extract the corresponding data from the manufacturer specification and tag it with a unique key associated with the user manual, and (i) once the system details are in place, use the maintenance manual to merge the maintenance operations involving the specific system configuration and tag them with the same unique ID. After completing the above three steps, the data will be labeled for each system model. A mapping process is then performed in which each fault known to the system is extracted and mapped to the errors recorded in the system log and correlations are extracted from these errors to predict the nature of the data. Figure 4 The process is shown diagrammatically.

[0042] refer to Figure 5 , it is possible to use unstructured data tables to represent components and extract the operating parameters and their possible values ​​for each component using CER 32, such as Figure 5 As shown. Let P1, P2, P3, ..., P N are N operating parameters extracted from a fixed number of components, and let Vk1, Vk2, Vk3, ..., Vk Nrepresents the possible parameter values ​​for the corresponding components and parameters. These are the set of parameters extracted from all manuals using NLP 30 and CER 32. The goal is to find meaningful combinations of these parameters and their corresponding optimal limits (Ps1, Ps2, Ps3, ..., Ps M , M≤N), thus generating model M c to generate a good pool or a bad pool on a per-component basis. This can be explained using the following notation:

[0043] M2=∪P i is a set of rules based on only one parameter;

[0044] It is a set of rules based on each tuple combination;

[0045] It is a set of rules based on k-tuples;

[0046] but

[0047] make Is to use all M k The optimal combination of rules is considered in the model.

[0048] refer to Figure 6 The resulting output of the knowledge engine 22 will include Figure 6 A set of decision rules 36 for each component, as depicted in the sample table of FIG. , helps to classify the data into good pools and bad pools. Figure 6 and Figure 3 The comparison illustrates how the decision rule can be formulated as Figure 6 The maintenance recommendations are shown in the rightmost column.

[0049] In the following, it is described Figure 1 Some illustrative implementations of ETL 42 and applications of knowledge engine 22 are shown.

[0050] The mapping process between the knowledge engine 22 and the machine log data 40 involves parsing the machine log data into a system readable format and storing it in the database system. Figure 1 The ETL 42 indicated in FIG. 40. The log data stored in the database will usually be distributed in multiple tables of the database. The first step to achieve the goal of mapping the knowledge engine 22 with the machine log 40 is to aggregate the related log data from the database. The log data can be related in two ways: (1) the time correlation of the log information; (2) sharing common components or subsystems.

[0051] refer to Figure 7,In the case of temporal correlation of log information, log data is extracted from the ,database, and the log data is aligned (increased / decreased) in the time domain, and temporal pattern matching is ,used to extract relevant log information, e.g. Figure 7 shown.

[0052] Various methods can be used to detect and aggregate log information belonging to components / subsystems of the same type to detect related log data. One suitable method is to perform lexical analysis on the content of the log information and correlate the lexical analysis outputs of two or more log instances. A threshold value on the correlation value is used to classify whether a given set of log information is related. Another illustrative method is to count the occurrence or non-occurrence of two or more error log data (usually data generated within a day) within a given time range and classify them as belonging to the same group or different groups. Yet another illustrative method is to calculate the mutual exclusivity between given sets of log information, and two or more log information data with the smallest mutual exclusivity value will be considered to belong to the same group of errors.

[0053] refer to Figure 8 Once the log data is divided into several groups, the matching of the system log data 40 with the knowledge engine 22 is performed. In order to perform the mapping 50 of the log data, each group of log information 52 in the grouped log information is sent to a splitter and lexical analyzer (SLA) unit 54. The SLA unit 54 first splits the log information into meaningful tokens (usually words), and then correlates each token with a set of words of a vocabulary 56 derived from the source (e.g., maintenance and operation data of electronic medical imaging equipment, such as explanatory user manuals 27, specification data 26, maintenance manuals 25) from which the knowledge engine is generated. Each element in the vocabulary 56 is assigned a weight based on its importance. The weight is determined by using domain expertise and / or by using statistical techniques (e.g., word counts that appear in the knowledge engine, the number of connections of a given word with unique different words, etc.). The vocabulary 56 is appropriately derived from the CER 32. The mapper 50 uses the weighted and labeled data from the SLA unit 54 to map the log information with entries (e.g., decision rules) of the knowledge engine 22, since at this stage both types of data (log data and knowledge engine rules) will be of the same nature and adopt a common language due to the fact that the log data is processed using a stored vocabulary 56. The mapping 50 can be achieved by techniques such as word similarity matching or associating words.

[0054] refer to Fig. 9 , which illustrates the Figure 1The method of appropriately executing the monitoring equipment to construct the decision rules 36 of the knowledge engine 22. In operation 60, various electronic medical imaging equipment maintenance and operation data 25, 26, 27, 28 are input to the knowledge engine builder 20. In operation 62, the reader module (e.g., the descriptive NLP 30 and the CER 32) is applied to extract the component ID and the operating parameters and the associated operating parameter ranges from the manual. In operation 64, the knowledge engine 22 is constructed by an operation including formulating the operating parameter range as a set of decision rules 36 for classifying the medical imaging equipment log data as good log data or bad log data. Fig. 9 As shown on the right side of FIG, during the deployment of the knowledge engine 22, the ETL 42 pre-processes the log files 40 from the medical imaging device 2. The CER 32 is applied to the ETL processed log data to extract the component ID, the operating parameters, and the values ​​of the operating parameters. This is suitably done by applying the same CER 32 used in the process manual to the ETL processed log data. In operation 66, the decision rules 36 of the knowledge engine 22 are mapped to the log files output by the ETL / CER processing chain. This can be done as already mentioned with reference to FIG. Figure 8 Operation 66 is performed as described above. In operation 68, the output of mapping 66 is used to classify the log data into good data (wherein the operating parameters are within the range defined by the decision rules 36) or bad data (wherein at least one operating parameter is outside the range defined by the decision rules 36). In operation 70, a maintenance alert 24 is issued based on the detected bad data. In an optional adaptive update operation 72, the bad data can be used to update the decision rules 36 based on whether the maintenance alert is determined to be accurate (e.g., based on information from the maintenance log 44 regarding whether the subject component was maintained or failed during the MTTF time interval after the maintenance alert was issued).

[0055] refer to Fig.10 and Fig.11 , describes an illustrative more detailed implementation of the read operation 62 (eg, an illustrative NLP 30 and CER 32). Fig.10 The text content of the original manual data 80 generally includes tables and text (e.g., paragraphs, sentences, bullets, ordered lists, etc.). The table parser 82 parses the table content into semantic units such as rows and / or cells, while the text parser 84 parses the text content into semantic units such as sentences or paragraphs. NLP 30 and CER 32 are then applied as described above to extract component IDs and operating parameters and associated operating parameter ranges or values. Fig.10In the example of , the latter processing is shown separately as a value engine component 86, where parameter value association is performed on a sentence-by-sentence, unit-by-unit or other semantic unit basis to generate a component-by-component output 88. Fig.11 , the assimilation unit 90 combines the component-by-component outputs 88 from the various electronic medical imaging equipment maintenance and operation data 25, 26, 27, 28 to generate a combined unstructured data table 92, as already described with reference to Figure 5 Then, the data table 92 is used as a Fig. 9 The knowledge engine builds the input of operation 64.

[0056] The present invention has been described with reference to the preferred embodiments. Modifications and substitutions may occur to others upon reading and understanding the foregoing detailed description. It is intended that the exemplary embodiments be interpreted as including all such modifications and substitutions as long as they fall within the scope of the claims and their equivalents.

Claims

1. A non-transitory storage medium (12) storing instructions, the instructions being readable and executable by an electronic processor (10, 16) to perform a monitoring method for generating a maintenance alert, the monitoring method comprising: extracting a component identification identifying a medical imaging device component and an operating parameter of the identified medical imaging device component and an associated operating parameter range from electronic medical imaging device maintenance and operation data (25, 26, 27, 28) including at least one electronic medical imaging device manual; generating a knowledge engine (22) by an operation including formulating the operating parameter range as a set of decision rules (36) for classifying medical imaging device log data, wherein the set of decision rules in the knowledge engine is formulated based solely on the electronic medical imaging device maintenance and operation data including the at least one electronic medical imaging device manual; and The knowledge engine is applied to log data (40) generated by a monitored medical imaging device (2) to detect out-of-range log data generated by the monitored medical imaging device and generate a maintenance alert (24) in response to the detected out-of-range log data, wherein the maintenance alert is associated with a component identification contained in the detected bad log data.

2. The non-transitory storage medium (12) according to claim 1, wherein: Formulating the operating parameter range as the set of decision rules (36) includes formulating a decision rule to classify the input log data as out-of-range log data if the input log data indicates a value for the operating parameter that is outside the operating parameter range associated with the operating parameter.

3. The non-transitory storage medium (12) according to any one of claims 1 to 2, wherein: The monitoring method further comprises: determining the accuracy of the generated maintenance alert based on feedback extracted from a maintenance log (44) of the monitored medical imaging device (2); and The knowledge engine (22) is updated by adjusting the decision rules (36) based on the determined accuracy of the generated maintenance alert (24).

4. The non-transitory storage medium (12) according to claim 3, wherein: Applying the knowledge engine includes generating a maintenance alert (24) in response to detecting a combination of two or more parameters in the out-of-range log data that are outside of their respective operating parameter ranges.

5. The non-transitory storage medium (12) according to any one of claims 1, 2 and 4, wherein: The named entity recognition algorithm is used to extract component identifiers.

6. The non-transitory storage medium (12) according to claim 5, wherein: Extracting operating parameters of identified medical imaging device components and associated operating parameter ranges includes: Parsing the text and tables of the electronic medical imaging equipment manual (25, 26, 27, 28) to delineate semantic units including at least sentences, list items, and table rows or cells; identifying an operating parameter by identifying a numerical value and a parameter item connected by a link item or symbol indicating equality or inequality, and associating the parameter identified by the parameter item with a medical imaging device component whose component identification appears in the same semantic unit as the parameter item; and An operating parameter range associated with an operating parameter is determined based on the numerical value and the equality or inequality indicated by the linking term or symbol.

7. The non-transitory storage medium (12) according to any one of claims 1, 2, 4 and 6, wherein: The generated maintenance alert (24) is formulated as a maintenance recommendation, and the monitoring method further includes controlling a display (18) to present the maintenance recommendation in response to the detected out-of-range log data.

8. A monitoring device for generating a maintenance alarm, the monitoring device comprising: Display (18); A non-transitory storage medium (12) according to any one of claims 1 to 7, which can be read and executed by an electronic processor (10, 16) to perform a monitoring method for generating a maintenance alarm; as well as An electronic processor (10, 16) is operatively connected to the display and the non-transitory storage medium to execute the monitoring method, the monitoring method also comprising displaying the generated maintenance alert (24) on the display.

9. The monitoring device according to claim 8, wherein: The monitoring method further comprises: determining the accuracy of a presented maintenance alert (24) based on feedback extracted from a maintenance log (44) of the monitored medical imaging device (2); and The decision rule (36) is updated based on the determined accuracy of the presented maintenance alert.

10. The monitoring device according to claim 9, wherein: The accuracy of the presented maintenance alert (24) is determined based on feedback extracted from the maintenance log (44) of the monitored medical imaging device (2) over a time interval extending from the time of presentation of the maintenance alert to an end time dependent on a mean time between failures of a component of the medical imaging device identified by the component identification associated with the maintenance alert.

11. A monitoring method for generating a maintenance alarm, the monitoring method comprising: extracting, by an electronic processor (10, 16), a component identification identifying a medical imaging device component in an electronic medical imaging device manual (25, 26, 27, 28); extracting, by the electronic processor, operating parameters and associated operating parameter ranges for the identified medical imaging device component from the electronic medical imaging device manual based on identifying numerical values ​​of operating parameters, parameter terms, and linking terms or symbols connecting the numerical values ​​and parameter terms indicating equality or inequality; formulating, by the electronic processor, the operating parameter range as a decision rule (36) for classifying medical imaging device log data based on whether the value of the associated operating parameter is outside the operating parameter range, wherein the decision rule is formulated based solely on the electronic medical imaging device manual; and applying, by the electronic processor, the decision rule to log data (40) generated by the monitored medical imaging device (2) to detect out-of-range log data generated by the monitored medical imaging device; and A maintenance alert (24) is displayed on a display (18) in response to detecting out-of-range log data, wherein the maintenance alert is generated based on the out-of-range log data and is associated with a component identification contained in the out-of-range log data.

12. The monitoring method according to claim 11, wherein: The monitoring method further comprises: determining, by the electronic processor (10, 16), the accuracy of a maintenance alert (24) based on feedback extracted from a maintenance log (44) of the monitored medical imaging device (2); and The decision rule is updated based on the determined accuracy (36).

13. The monitoring method according to any one of claims 11 to 12, wherein: Applying the decision rule to detect out-of-range log data includes applying the decision rule to detect a combination of two or more parameters in the out-of-range log data that is outside of their respective operating parameter ranges.

14. The monitoring method according to any one of claims 11 to 12, wherein: The maintenance alert (24) is formulated as a recommendation for checking the value of the operating parameter associated with the operating parameter range of the decision rule (36) that generated the maintenance alert.

15. The monitoring method according to any one of claims 11 to 12, wherein: The extraction is performed on an electronic medical imaging device manual (25, 26, 27, 28) including at least one of a user manual (27) or an operating reference manual (28).

Images