JP7358175B2 - Diagnostic support device and diagnostic support program - Google Patents

Description

The present embodiment relates to a diagnosis support device and a diagnosis support program.

2. Description of the Related Art In recent years, there has been a demand for a system that displays on a single screen various diagnostic data necessary for doctors to consider diagnosis and treatment strategies. For example, it is possible to simultaneously view and compare information on treatments such as medication, information on tests such as laboratory tests and image tests, etc., and to understand the condition of the subject and current treatment details with a high level of visibility. Equipment and systems are being developed.

When making medical decisions such as early detection of deterioration of a condition using such devices, it is generally necessary to measure changes over time in diagnostic data (e.g., physiological data), and calculate changes from the previous measurement value. Techniques such as issuing an alert when the rate exceeds a threshold have been adopted.

On the other hand, when making medical decisions using diagnostic data, there are often cases in which changes in diagnostic data over time are important. For example, renal dysfunction (WRF) that occurs during the treatment of heart failure is defined as "an increase in serum Cr of 0.3 mg/dL or more within 48 hours" and is used as a standard for medical judgment.

JP2018-47262A

When detecting changes in diagnostic data values within a certain period of time, there are various patterns of changes. Diagnosis and treatment responses may also change depending on the pattern. Generally, these cannot be distinguished by a simple threshold value. For this reason, diagnostic support through detection processing using only thresholds may lead to oversights or misjudgments. Therefore, new support technology is desired for detecting such change patterns in diagnostic data.

An object of the present invention is to provide a diagnosis support device and a diagnosis support program that can perform detection according to various change patterns in diagnostic data.

The diagnosis support device according to this embodiment includes an input section, a representative pattern determination section, a detection algorithm determination section, and a detection section. The input unit inputs a case to be detected. The representative pattern determination unit determines, for the selected case, a first representative pattern regarding a temporal change aspect of at least the first diagnostic data. The detection algorithm determination unit determines a first detection algorithm according to the determined first representative pattern. The detection unit detects the onset of the case based on the determined first detection algorithm.

FIG. 1 is a diagram showing an example of the configuration of a diagnostic support system S according to the first embodiment. FIG. 2 is a block diagram showing the configuration of the diagnostic support device 1. As shown in FIG. FIG. 3 is a diagram showing an example of the registration screen (registration GUI) 31 used in the diagnostic support process in the registration mode. FIG. 4 is a flowchart showing the flow of diagnostic support processing in the registration mode executed by the diagnostic support device 1. FIG. 5 is a diagram showing an example of a registration screen when a rule name is input in registration mode. FIG. 6 is a diagram showing an example of a registration screen when characteristics of a representative pattern in text are input in the registration mode. FIG. 7 is a diagram showing an example of a registration screen when characteristics of a representative pattern based on a graph are input in the registration mode. FIG. 8 is a diagram showing an example of a registration screen when characteristics of a representative pattern based on a graph are input in the registration mode. FIG. 9 is a diagram showing an example of a registration screen when characteristics of a representative pattern based on a graph are input in the registration mode. FIG. 10 is a diagram for explaining a detection algorithm determined according to registered representative patterns. FIG. 11 is a flowchart showing the flow of diagnostic support processing in the detection mode executed by the diagnostic support device 1. FIG. 12 is a flowchart showing the flow of diagnostic support processing in the verification mode executed by the diagnostic support device 1. FIG. 13 is a diagram for explaining diagnosis support processing according to modification example 4, which is executed by the diagnosis support apparatus 1. FIG. 14 is a diagram for explaining diagnosis support processing according to modification 5, which is executed by the diagnosis support apparatus 1. FIG. 15 is a diagram for explaining diagnosis support processing according to modification 6, which is executed by the diagnosis support apparatus 1.

Hereinafter, embodiments will be described according to the drawings. In the following description, components having substantially the same functions and configurations will be denoted by the same reference numerals, and duplicate description will be given only when necessary. Furthermore, the embodiments can be combined with other configurations or conventional techniques as long as there is no inconsistency in the configuration.

FIG. 1 is a diagram showing a configuration example of a diagnostic support device system S according to an embodiment.

For example, as shown in FIG. 1, the diagnosis support system S according to the present embodiment includes a diagnosis support device 1, a medical image diagnosis device 3, an HIS (Hospital Information System) 5, an RIS (Radiology Information System) 7, and a PACS (Picture System). Archiving and Communication Systems)9. Each device is installed in a hospital, for example, and can communicate with other devices via a network N such as an in-hospital LAN (Local Area Network).

The diagnosis support device 1 executes diagnosis support processing. The diagnosis support device 1 acquires various diagnostic data from the HIS 5, RIS 7, PACS 9, and the medical image diagnostic device 3 via the network N, and performs various information processing using the acquired diagnostic data. For example, the diagnosis support device 1 is realized by computer equipment such as a workstation.

The medical image diagnostic device 3 is a device that generates medical images based on data collected from a subject. For example, the medical image diagnostic apparatus 3 includes an X-ray diagnostic apparatus, an X-ray CT (Computed Tomography) apparatus, an MRI (Magnetic Resonance Imaging) apparatus, an ultrasound diagnostic apparatus, a SPECT (Single Photon Emission Computed Tomography) apparatus, a PET (Positron Emission These include a SPECT-CT device in which a SPECT device and an X-ray CT device are integrated, a PET-CT device in which a PET device and an X-ray CT device are integrated, and the like.

FIG. 2 is a block diagram showing the configuration of the diagnostic support device 1. As shown in FIG. As shown in the figure, the diagnostic support device 1 includes an I/F (interface) circuit 10, an input circuit 20, a display 30, a storage circuit 40, and a processing circuit 50.

The I/F circuit 10 is connected to the processing circuit 50 and controls transmission and communication of various data between the medical image diagnostic apparatus 3, HIS 5, RIS 7, and PACS 9. For example, the I/F circuit 10 receives diagnostic data from the medical image diagnostic apparatus 3, HIS 5, RIS 7, and PACS 9, and outputs the received diagnostic data to the processing circuit 50. For example, the I/F circuit 10 is realized by a network card, a network adapter, a NIC (Network Interface Controller), or the like.

The input circuit 20 is connected to the processing circuit 50 , converts an input operation received from an operator into an electrical signal, and outputs the electrical signal to the processing circuit 50 . For example, the input circuit 20 is realized by a trackball, a switch button, a mouse, a keyboard, a touch panel, or the like.

The display 30 is connected to the processing circuit 50 and displays various information and various image data output from the processing circuit 50. For example, the display 30 is realized by a liquid crystal monitor, a CRT (cathode ray tube) monitor, a touch panel, or the like.

The storage circuit 40 is connected to the processing circuit 50 and stores various data. For example, the storage circuit 40 includes diagnostic data received from the medical image diagnostic apparatus 3, HIS 5, RIS 7, and PACS 9, a detection algorithm table for managing a detection algorithm described later, a dedicated program for realizing diagnosis support processing described below, etc. remember. For example, the memory circuit 40 is realized by a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory, a hard disk, an optical disk, or the like.

The processing circuit 50 is a processor serving as a CPU that controls the entire processing of the diagnostic support device 1 . For example, the processing circuit 50 controls the components of the diagnostic support device 1 according to input operations received from the operator via the input circuit 30. Further, for example, the processing circuit 50 controls the display 30 to display diagnostic data stored in the storage circuit 40, a GUI for performing various input processes, arithmetic processing results, and the like.

Furthermore, the processing circuit 50 has a representative pattern determining function 50a, a detection algorithm determining function 50b, a verification processing function 50c, and a detection processing function 50d.

The representative pattern determination function 50a determines, for the selected case, a representative pattern regarding at least the temporal change aspect of the diagnostic data.

Here, the "representative pattern for a case" refers to what diagnostic data and from what perspective the values and changes in the data can be grasped to determine whether the case has symptoms or has developed symptoms. The standard is arbitrarily defined by the user for each case as a typical pattern. Note that the diagnostic data refers to physiological data of a patient, such as body temperature, blood pressure, respiratory rate, heart rate, serum value, etc. For example, a typical pattern of renal dysfunction (WRF) that occurs during the treatment of heart failure can be defined as "an increase in serum Cr of 0.3 mg/dL or more within 48 hours."

The detection algorithm determination function 50b determines a detection algorithm according to the determined representative pattern.

Here, the "detection algorithm" is an algorithm for data analysis that is determined in advance according to the representative pattern. For example, with respect to changes in body temperature over time, it is an algorithm for data analysis that detects the "difference between the start and end points (of the analysis period)," the "average error from a linear change," or a combination thereof.

The verification processing function 50c verifies whether or not the determined representative pattern and detection algorithm are appropriate for detecting the target case.

The detection processing function 50d detects the onset of the case based on the detection algorithm determined by the detection algorithm determination function 50b. Specifically, the detection processing function 50d extracts target diagnostic data using a representative pattern determined for each case, and extracts the extracted diagnostic data using a detection algorithm determined according to the representative pattern. Analyze and detect data applicable to the case. That is, the detection processing function 50d extracts the corresponding diagnostic data using the representative pattern registered by the representative pattern determination function 50a. Furthermore, the extracted diagnostic data is analyzed using a detection algorithm associated with the registered representative pattern, and data corresponding to the case is detected to support diagnosis.

Note that the representative pattern determining function 50a, the detection algorithm determining function 50b, the verification processing function 50c, and the detection processing function 50d are realized by the control circuit 16 serving as a CPU executing a control program. However, the present invention is not limited to this example, and part or all of the image generation function 16a and the image processing function 16b may be implemented using dedicated hardware designed to perform similar functions, such as an ASIC (Application Specific Integrated Circuit). , DSP (digital signal processor), FPGA (field programmable gate array), or other semiconductor integrated circuits or conventional circuit modules may be used.

(Diagnosis support processing using representative patterns for each case)
Next, a diagnosis support process executed by the diagnosis support apparatus according to this embodiment will be described. This diagnosis support process can be classified into diagnosis support process in "registration mode", diagnosis support process in "detection mode", and diagnosis support process in "verification mode". The diagnosis support processing according to each mode will be explained below.

(Diagnosis support processing in registration mode)
The diagnosis support process in the "registration mode" determines a first representative pattern regarding the temporal change aspect of at least the first diagnostic data for the selected case, and detects the detected case according to the determined first representative pattern. This process includes at least determining an algorithm. Diagnosis support processing in this "registration mode" is executed by the representative pattern determination function 50a and the detection algorithm determination function 50b.

FIG. 3 is a diagram showing an example of the registration screen (registration GUI) 31 used in the diagnostic support process in the registration mode. This registration screen is displayed on the display 30, for example, when the registration mode is activated. Note that the registration screen shown in FIG. 3 is just an example. That is, any registration screen may be used as long as it is for registering at least a representative pattern regarding the temporal change aspect of diagnostic data.

As shown in the figure, the registration screen has a rule name input area 32, a text property input area 33, a graph property input area 34, and an operation button area 35. Note that "characteristics" in the registration screen means characteristics of the representative pattern, and particularly includes at least the aspect of temporal change of the first diagnostic data. However, not only the temporal change aspect of the first diagnostic data, but also the temporal change aspect of the second diagnostic data different from the first diagnostic data, information such as threshold values, etc. "Characteristics". In the following explanation, when handling multiple diagnostic data, they will be referred to as "first diagnostic data", "second diagnostic data", etc. with ordinal numbers added to distinguish them. shall be. Furthermore, when handling one type of diagnostic data, it may be referred to simply as "first diagnostic data" and "diagnostic data."

The rule name input area 32 is an area provided with a GUI for inputting a rule name (case name). The user inputs the rule name by directly inputting the case name into the box provided in the rule name input area 32 via the input circuit 20, or by selecting and inputting the desired case name from the pull-down menu. can be entered.

The text property input area 33 is an area provided with a GUI for inputting the properties of the representative pattern by text input. The user can input the representative pattern by directly inputting the necessary information into each box provided in the text-based characteristic input area 33 via the input circuit 20, or by selecting and inputting the desired characteristic from the pull-down menu. properties can be defined textually.

Note that, in accordance with the case name input in the rule name input area 32, a characteristic pattern prepared in advance may be automatically displayed as a candidate pattern in the text characteristic input area 33. Here, the "candidate pattern" is typically a pattern candidate prepared in advance for the input rule (case) based on, for example, a clinical practice guideline or a representative pattern registered in the past.

For example, by using this candidate pattern, if you enter "sepsis" in the rule name input area 32, in view of the symptom of a rise in body temperature over a certain period of time, you can select, for example, "Your body temperature will rise by 2 degrees Celsius within 48 hours." Candidate patterns using text (characters, sentences) such as "I do" can be automatically displayed in the text characteristic input area 33. By appropriately modifying the displayed candidate pattern, the user can determine the characteristic of the representative pattern (that is, the temporal change mode of the diagnostic data) desired by the user using the text.

The graphical characteristic input area 34 is an area provided with a GUI for inputting (defining) the characteristic of the representative pattern using a graph. For example, in the graph characteristic input area 34, when the characteristics of a representative pattern are defined by text input in the text characteristic input area 33, a plurality of graph change patterns corresponding to the contents are automatically created. Is displayed. The user selects a desired change pattern from among the displayed graph patterns via the input circuit 20, thereby determining the characteristics of the representative pattern (i.e., the mode of change over time of the diagnostic data). can be defined.

Note that the graph pattern displayed in the graph characteristic input area 34 does not necessarily need to be linked with the text characteristic input area 33.

The operation button area 35 is an area provided with a GUI for inputting instructions using four buttons: "Specify change pattern," "Save," "Cancel," and "Verify." “Specify change pattern” is a button for inputting an instruction to display a graph pattern in the graph characteristic input area 34. "Save" is a button for instructing to save (register) a representative pattern whose characteristics are defined by text and graphs. "Cancel" is a button for resetting the information input in each of the rule name input area 32, text property input area 33, and graph property input area 34. “Verification” is a button for instructing activation of a verification mode, which will be described later.

Next, the flow of diagnostic support processing in registration mode will be explained.

FIG. 4 is a flowchart showing the flow of diagnostic support processing in the registration mode executed by the diagnostic support device 1.

As shown in the figure, when predetermined information (user ID, password, etc.) is input via the input circuit 20, the user is logged into the diagnostic support device 1 (step S1). Further, when the "registration mode" is selected via the input circuit 20, the processing circuit 50 develops the dedicated program stored in the storage circuit 40 on the memory, thereby implementing the representative pattern determination function 50a, the detection algorithm The decision function 50b is implemented and the registration mode is activated. The representative pattern determination function 50a displays a registration screen on the display 30 (step S2).

The representative pattern determination function 50a receives the case name input into the box of the rule name input area 32 via the input circuit 10 (step S3). In this embodiment, in order to make the explanation concrete, it is assumed that "sepsis" is input as the case name, as shown in FIG. 5. Note that inputting the case name also includes a selection operation from a plurality of preset cases. Further, although it is typically assumed that the input is performed manually by a person (user), the input may be automatically input based on the information input in step S1, for example.

The representative pattern determination function 50a displays a text candidate pattern in the text characteristic input area 33 according to the rule name (case name; in this case, "sepsis") input in step S4. For example, a text candidate pattern is presented in which "body temperature" is defined as "increase" by "2.0 degrees Celsius" within "48 hours" (step S4).

The user changes the presented text candidate pattern by direct input or selection input from a pull-down menu, as necessary. In this embodiment, as shown in FIG. 6, the text is corrected to "body temperature will rise by 2.0° C. within 72 hours", and the characteristics of the representative pattern are determined by text (step S5).

Once the characteristics of the representative pattern based on the text are determined, the user presses the "Specify change pattern" button in the operation button area 35. The representative pattern determination function 50a, triggered by the pressing of "Specify change pattern", selects sepsis as a characteristic of the representative pattern defined by the text as "body temperature rises by 2.0 degrees Celsius within 72 hours". In order to further define the aspect of temporal change using a graph, at least one (graph) change pattern is displayed in the graph characteristic input area 34 (step S6).

7, FIG. 8, and FIG. 9 are diagrams showing an example of a registration screen in which a graph of a change pattern is displayed in the graph characteristic input area 34. Hereinafter, a case will be described in which the characteristic of a representative pattern is defined in the characteristic input area 34 using a graph, using changes in the graph from three viewpoints: "slope pattern", "change period", and "irregular fluctuation".

For example, as shown in FIG. 7, the representative pattern determining function 50a determines a "linearly changing pattern" and a "pulsing pattern" in the graph characteristic input area 34 in order to determine a "sloping pattern" as a first step. A graph corresponding to each change pattern is displayed. When a graph corresponding to one of the slope patterns is selected by the user via the input circuit 20 (the figure exemplifies the case where the "Linearly changing pattern" checkbox is checked), The representative pattern determination function 50a determines the slope pattern corresponding to the selected graph as the slope pattern of the representative pattern (step S7).

Further, the representative pattern determination function 50a displays the slope patterns not selected in the first step in a blind display, a hatched display, an inverted display, etc. so as not to attract the user's attention.

When the selection in the first step is completed, the representative pattern determination function 50a, as shown in FIG. Graphs corresponding to respective change patterns are displayed: a "change period: long" pattern in which body temperature changes over a long period of time, and a "change period: short" pattern in which body temperature changes over a relatively short period of time. Note that variations in the pattern of change periods displayed in the second step depend on which slope pattern was selected in the first step. The example shown in the figure illustrates a case where, as a result of selecting the "linearly changing pattern" in the first step, a "change period: long" pattern and a "change period: short" pattern are displayed.

When a graph corresponding to one of the change period patterns is selected by the user via the input circuit 20 (the figure exemplifies the case where the "change period: long" checkbox is checked), The representative pattern determination function 50a determines the change period pattern corresponding to the selected graph as the change period pattern of the representative pattern (step S8).

Further, the representative pattern determination function 50a displays the change period patterns that were not selected in the second step in a blind display, a shaded display, an inverted display, etc. so as not to draw the user's attention, as in the case of the first step. do.

When the selection in the second step is completed, the representative pattern determination function 50a, as shown in FIG. Graphs corresponding to the respective change patterns of the "irregular fluctuation: small" pattern where the irregular fluctuation is relatively small over the period of change and the "irregular fluctuation: large" pattern where the irregular fluctuation is relatively large over the change period are shown. indicate. Note that variations in the changing period pattern presented in the third step also depend on which changing period pattern was selected in the second step. In the same figure, the "irregular fluctuation: small" pattern and the "irregular fluctuation: large" pattern are exemplified when the change period is long as a result of selecting "change period: long" in the second step. ing.

A graph corresponding to one of the irregular fluctuation patterns is selected by the user via the input circuit 20 (the figure exemplifies the case where the check box for the "irregular fluctuation: small" pattern is checked). ), the representative pattern determination function 50a determines the irregular fluctuation pattern corresponding to the selected graph as the irregular fluctuation pattern of the representative pattern (step S9).

Once the characteristics of the representative pattern are determined based on the text and graph, the user operates the save button 35b in the operation button area 35 to instruct registration (save) of the determined representative pattern. The representative pattern determination function 50a registers the determined representative pattern as a representative pattern of "sepsis" in response to the operation of the save button 35b (step S10).

Note that when the representative pattern of "sepsis" is saved in step S10, meta information including the user ID and save time input in step S1 is saved in association with the representative pattern.

The detection algorithm determination function 50b determines a detection algorithm for "sepsis" based on the registered representative pattern. That is, the detection algorithm determination function 50b determines a detection algorithm regarding "sepsis" by referring to the detection algorithm table stored in advance in the storage circuit 40 and the registered representative pattern (step S11).

That is, the representative patterns registered as "sepsis" in step S10 are "body temperature rises by 2.0°C within 72 hours", "slope pattern", "change period: long", and "irregular fluctuation: small". It has the following characteristics. The detection algorithm determination function 50b extracts an algorithm for detecting body temperature changes within 72 hours based on the characteristic that "body temperature increases by 2.0° C. within 72 hours." As a result, as shown in FIG. A case is assumed in which four detection algorithms are extracted: "difference between the maximum value and the maximum value" and "average error from linear change within 72 hours." The detection algorithm determination function 50b selects one of the above four detection algorithms based on the fact that the registered representative pattern further has the characteristics of "slope pattern", "change period: long", and "irregular fluctuation: small". The corresponding detection algorithm is extracted from

As a result, the detection algorithm determination function 50b extracts two detection algorithms: "difference between starting point and maximum value within 72 hours" and "average error from linear change within 72 hours", for example. The detection algorithm determining function 50b registers (saves) the determined algorithm as a detection algorithm for "sepsis" in association with the representative pattern (step S12).

Note that when the "sepsis" detection algorithm is stored in association with the representative pattern in step S12, the meta information including the user ID and storage time input in step S1 is stored in association with the representative pattern.

(Diagnosis support processing in detection mode)
Next, the flow of diagnostic support processing in detection mode will be explained. In the following, in order to make the explanation more concrete, a case will be exemplified in which a representative pattern for "sepsis" registered in the diagnosis support process in the registration mode described above is used.

In addition, the diagnostic support processing in this detection mode can be used, for example, when detecting signs of sepsis or the presence or absence of the onset of sepsis using diagnostic data acquired in the past for a specific patient, for example, using a wearable device, etc. When detecting the presence or absence of signs or onset of sepsis from diagnostic data obtained in real time using diagnostic data, or identifying patients with signs or onset of sepsis from diagnostic data of multiple patients, etc.

FIG. 11 is a flowchart showing the flow of diagnostic support processing in the detection mode executed by the diagnostic support device 1.

As shown in the figure, when predetermined information (user ID, password, etc.) is input via the input circuit 20, the user is logged into the diagnostic support device 1 (step S21). Furthermore, when the "detection mode" is selected and input through a predetermined operation via the input circuit 20, the processing circuit 50 develops the dedicated program stored in the storage circuit 40 onto the memory. As a result, the detection processing function 50b is realized and the detection mode is activated (step S2).

The detection processing function 50b reads out the representative pattern associated with the rule name (case name) input by the user from the storage circuit 40 (step 23). Note that when a plurality of representative patterns are associated with a rule name, the detection processing function 50b reads out one of the representative patterns selected and input by the user.

The detection processing function 50b executes detection processing using the read representative pattern and the detection algorithm associated with the representative pattern. For example, when detecting the presence or absence of the onset of sepsis for a specific patient using diagnostic data acquired in the past, a representative pattern read out from the diagnostic data acquired in the past (or Extract diagnostic data (meeting the conditions). The detection processing function 50b uses a detection algorithm associated with a representative pattern on the extracted diagnostic data to determine the presence or absence of signs and onset of sepsis, and performs a detection process to identify the diagnostic data that serves as the basis. Execute (step S24).

The detection processing function 50b displays the results of the detection processing (for example, information regarding the presence or absence of signs and onset of sepsis, the diagnostic data used as the basis, and the representative pattern) on the display 30 in a predetermined format as diagnostic support information.

(Diagnosis support processing in verification mode)
Next, the flow of diagnostic support processing in verification mode will be explained. Note that verification using this verification mode can be performed at any timing. Typically, before the representative pattern determined in the registration mode is saved (registered), the registered representative pattern can be verified later at any timing.

FIG. 12 is a flowchart showing the flow of diagnostic support processing in the verification mode executed by the diagnostic support device 1.

As shown in the figure, when the "Verify" button is operated via the input circuit 20 to instruct activation of the verification mode, the processing circuit 50 expands the dedicated program stored in the storage circuit 40 onto the memory. As a result, the verification processing function 50c is realized and the verification mode is activated (step S31). Note that the startup timing of this verification mode (that is, the timing of executing the diagnostic support process in the verification mode) is the same as step S11 and step S11 shown in FIG. If the verification is to be performed after registering the representative pattern and the detection algorithm, it will be after step S12 shown in FIG. 4.

The verification processing function 50c selects a database to be used for verification according to instructions from the user via the input circuit 20 (step S32).

Database selection is performed automatically or manually depending on the purpose of verification. For example, if the purpose is to verify the representative pattern and the accuracy of the detection algorithm, a first database is selected in which diagnostic data that is the basis for diagnosing sepsis for a plurality of patients is collected. In addition, if the purpose is to verify the possibility that the representative pattern and detection algorithm will detect other cases that are not sepsis, diagnostic data for multiple patients that is the basis for cases that are completely unrelated to sepsis should be used. A second database in which the information is collected is selected. Furthermore, if these two verifications are to be achieved simultaneously, the diagnostic data that served as the basis for the diagnosis of sepsis for multiple patients and the diagnostic data that served as the basis for the diagnosis of cases completely unrelated to sepsis will be collected. A third (mixed) database is selected.

The verification processing function 50c executes a detection process on diagnostic data included in the selected database using a representative pattern and a detection algorithm. The verification processing function 50c executes verification processing by calculating the matching rate of the detection results using the correct answers prepared in advance (step S33). For example, if the first database is selected in step S32, the verification processing function 50c calculates the correct answer rate from the number detected by the detection processing.

The verification processing function 50c displays the results of the verification processing (ie, match rate, correct answer rate) on the display 30 in a predetermined format.

By using the verification mode described above, it is possible to use objective data to determine whether the representative pattern that the user has determined and the detection algorithm that is based on the representative pattern are appropriate for detecting the target case. It can be verified by The user can determine a more accurate representative pattern and a detection algorithm based on the verification result in the verification mode. Furthermore, by verifying a plurality of representative patterns for a specific case, it is also possible to quantitatively verify which representative pattern is the most appropriate.

(effect)
According to the diagnosis support device 1 according to the present embodiment described above, the representative pattern determination function 50a determines, for the input case and diagnostic data, a representative pattern regarding at least the temporal change aspect of the diagnostic data, and detects The algorithm determining unit 50b determines a detection algorithm according to the determined representative pattern. The detection processing function 50d detects the onset of the case based on the determined detection algorithm. That is, the representative pattern determination function 50a can determine a representative pattern regarding the mode of change over time in diagnostic data according to the user's input (according to the user's intention), and can perform detection according to various change patterns in the diagnostic data. enables diagnostic support. By arbitrarily determining the mode of temporal change in diagnostic data, the user can extract diagnostic data that has the desired change pattern, and also create a detection algorithm that matches the representative pattern intended by the user. Using this, it is possible to detect the presence or absence of symptoms and the presence or absence of onset of symptoms in the target case.

Therefore, by utilizing the fact that changes and rates of change in diagnostic data differ depending on the case, it is possible to detect the presence or absence of symptoms and the presence or absence of onset of symptoms in a target case with a wide degree of freedom that reflects the user's intention. As a result, more advanced diagnostic support can be achieved than in the past.

Further, according to the diagnosis support device 1 according to the present embodiment, the representative pattern determining function 50a determines a representative pattern determined based on input from the user. Therefore, the user can freely determine the representative pattern with arbitrary contents, for example, by inputting text. Further, the representative pattern determination function 50a presents a plurality of graph change patterns classified in advance, and selects a representative pattern based on at least one graph change pattern selected from among the presented plurality of graph change patterns. Determine. Therefore, the user can arbitrarily determine a representative pattern by, for example, interactively selecting a desired pattern from a plurality of graph change patterns presented using a dedicated GUI. As a result, it is possible to easily and quickly register representative patterns.

(Modification 1)
In the embodiment described above, a configuration is exemplified in which both text and graph input are used in determining the characteristics of the representative pattern. However, both textual and graphical inputs are not essential in determining the characteristics of the representative pattern. That is, the characteristic of the representative pattern may be determined by either text or graph.
Furthermore, in the embodiment described above, a configuration is exemplified in which selection input using a graph is performed after input using text in determining characteristics of a representative pattern. However, the present invention is not limited to this example, and a configuration may be adopted in which, for example, a text input is performed after a graphic selection input. Furthermore, a configuration may be adopted in which text input and graph selection input are performed alternately.

(Modification 2)
In the above-mentioned embodiment, the case where the characteristic of the representative pattern using a graph is determined in three stages in the order of "inclined pattern", "change period", and "irregular fluctuation" has been exemplified. However, the present invention is not limited to this example, and may be defined in stages by including more variations as necessary.

Further, in the above embodiment, when inputting characteristics of a representative pattern using a graph, a case where only a "linearly changing pattern" is selected as a "slope pattern" is exemplified. However, as the "tilt pattern", it is also possible to select a "pattern that changes in a pulsed manner" in addition to a "pattern that changes linearly". In such a case, in the subsequent selection of "change period" and "irregular fluctuation", change patterns including variations according to "linear change pattern" and "pulse change pattern" will be displayed. It turns out.

(Modification 3)
In the embodiment described above, at least one graphical change pattern is displayed in the graphical characteristic input area 34, and the user selects a desired change pattern from among them, thereby inputting the aspect of temporal change using the graph. An example of this configuration has been explained. However, in order to realize more flexible input of the temporal change mode, the user is allowed to draw a graph change pattern via the input circuit 20, and based on the outline, the temporal change mode using the graph is determined. It may also be configured to input.

Note that the user can draw a graph change pattern by, for example, inputting an arbitrary graph outline into a drawing area provided in the graph characteristic input area 34 via the input circuit 20 (manual input), or by the same method. It is possible to adopt a configuration (semi-manual input) in which at least one graph change pattern candidate is displayed in the drawing area, and the user modifies and transforms the candidate to input the graph outline.

The representative pattern determination function 50a determines the mode of temporal change using a graph based on the similarity of the outline of the graph input by the user. Further, the detection algorithm determination function 50b determines a detection algorithm including an algorithm for calculating the similarity of the approximate shapes of the input graphs. For example, a DTW (Dynamic Time Warping) method can be used as an algorithm for calculating the similarity of the outlines of graphs. The detection processing function 50d calculates the similarity between the "representative pattern having the outline of a manually inputted or non-manually inputted graph" and the "change in actual diagnostic data" using the determined detection algorithm, and performs the detection processing. Execute. Note that the user may be able to specify a limit on the degree of temporal elasticity and a degree of similarity. As a result, detection processing with a higher degree of freedom can be performed using the DTW method that allows a certain error in shape.

(Modification 4)
In the embodiment described above, a case has been exemplified in which the representative pattern of a specific case (sepsis) is determined using text and graphs in terms of temporal changes in body temperature. Naturally, the representative pattern of a specific case may be determined by combining not only the temporal change of body temperature but also at least one other rule such as a threshold rule or a medication period.

FIG. 13 is a diagram for explaining diagnosis support processing in the registration mode according to the fourth modification. In the example shown in the figure, the representative pattern for a particular case is determined as "during drug A administration, test value X changes by more than β". When such a representative pattern is determined and registered, the detection processing function 50d specifies a period T1 during which drug A is being administered, and the value changes by more than β during this specified period T1 (diagnostic data Extract the graph of the test value X (as ). The detection processing function 50d displays the results of the detection processing (for example, the presence or absence of signs and onset of sepsis, the diagnostic data used as the basis, and information regarding the representative pattern) on the display 30 based on the graph of the extracted test value X. Displayed in the form of

(Modification 5)
In the above embodiment, the case is sepsis, and the diagnosis support process is illustrated in the registration mode in which only one type of diagnostic data (body temperature) is used to determine the characteristics of the representative pattern. However, the present invention is not limited to this example, and when determining a representative pattern for a particular case, two or more types of diagnostic data may be used to execute the diagnostic support process in the registration mode.

FIG. 14 is a diagram for explaining diagnosis support processing in the registration mode according to the fifth modification. In the example shown in the same figure, the representative pattern of a particular case uses test value Y as the first diagnostic data and test value X as the second diagnostic data, , the test value X changes by more than β.

When such a representative pattern is determined and registered, the detection processing function 50d specifies a period T2 in which the test value Y (as the first diagnostic data) is equal to or greater than α, and uses this specified period T2. , a graph of test values X (as second diagnostic data) whose values change by β or more is extracted. The detection processing function 50d displays the results of the detection processing (for example, the presence or absence of signs and onset of sepsis, the diagnostic data used as the basis, and information regarding the representative pattern) on the display 30 based on the graph of the extracted test value X. Displayed in the form of

(Modification 6)
The representative pattern determined according to the above modification 5 is based on a plurality of rules, if the characteristic using the first diagnostic data is the first rule, and the characteristic using the second diagnostic data is the second rule. It is possible to say things that are combined in a complex manner.
In this variation 6, detection processing is executed in parallel according to the plurality of rules, and if the result satisfies a certain condition, the detection processing is performed in parallel according to the plurality of rules. This method uses a representative pattern that satisfies the following criteria.

FIG. 15 is a diagram for explaining diagnosis support processing in the registration mode according to the sixth modification. In the figure, a characteristic using diagnostic data A is defined as rule A, a characteristic using diagnostic data B is defined as rule B, and a characteristic using diagnostic data C is defined as rule C. Furthermore, the hatched areas in the figure indicate the applicable period of each rule.

The representative pattern illustrated in the upper part of the figure is ``If three rules are detected in parallel as rule A, rule B, and rule C, and each is detected within a certain period of time, the three rules will be detected at the same time. It is assumed that it has been detected and is defined as "(Rule A) and (Rule B) and (Rule C)".

In such a case, if the simultaneity condition is "certain period = 24 hour unit", the detection processing function 50d extracts the graphs of diagnostic data A, B, and C corresponding to the period T3 shown in the middle part of FIG. . Further, if "certain period = 12 hour unit" is set, the detection processing function 50d executes the detection algorithm corresponding to each rule in parallel, and the diagnostic data A and B corresponding to the period T4 shown in the lower part of FIG. , extract the graph of C.

The detection processing function 50d uses the graphs of the extracted diagnostic data A, B, and C as a basis to determine the results of the detection process (for example, the presence or absence of signs and onset of sepsis, the diagnostic data used as the basis, and information regarding representative patterns). , displayed in a predetermined format on the display 30.

Note that the detection processing according to the present modification 6 can also be used, for example, when different characteristics are defined by a plurality of rules using the same diagnostic data and these are detected in combination. Even in this case, the detection processing in the detection mode is executed in parallel by each inspection algorithm corresponding to each rule, and diagnostic data that satisfies the concurrency condition is extracted.

(Modification 7)
In the embodiment described above, the configuration is such that the diagnosis support apparatus 1 executes the diagnosis support process in the detection mode. However, the diagnostic support process in the detection mode may be executed, for example, on a wearable device worn by the user, a mobile phone, or the like. Such a configuration can be realized by installing an application that realizes the detection processing function 50d in a wearable device or the like.

A user can launch an application on a wearable device or the like and execute the above-described diagnostic support process in the detection mode on diagnostic data acquired in real time. As a result of the detection process, if it is detected that ``the symptoms of the target case are present'' or ``the onset of the target case is present'', an alert, etc. (for example, ``There are signs of the onset of sepsis'', `` By notifying the user with a message such as "Please go to the hospital immediately", further early detection of cases and early countermeasures can be realized.

(Modification 8)
In the embodiment described above, when saving the determined representative pattern and the determined detection algorithm, meta information including the user ID and the storage time is stored in association with them. Using this meta information, it is also possible to manage representative patterns and detection algorithms for each case.

For example, for "sepsis," the determined representative patterns and determined detection algorithms can be displayed in a list for each user along with their storage times. Further, for a specific user, the determined representative pattern and the determined detection algorithm can be displayed in a list for each case along with their storage time.

By observing these list displays, the user can manage differences in representative patterns and detection algorithms between users. Therefore, for example, an inexperienced doctor can refer to representative patterns and detection algorithms created by veteran doctors, and can examine differences in representative patterns and detection algorithms between hospitals. Furthermore, the timing at which the representative pattern and detection algorithm were determined can be easily and quickly managed. Therefore, it is possible to determine whether or not the representative pattern and detection algorithm should be re-registered depending on the determined time.

Note that the term "processor" used in the above explanation refers to, for example, a CPU (central processing unit), a GPU (Graphics Processing Unit), an application specific integrated circuit (ASIC), or a programmable logic device. (For example, it refers to circuits such as Simple Programmable Logic Device (SPLD), Complex Programmable Logic Device (CPLD), and Field Programmable Gate Array (FPGA)). A processor realizes its functions by reading and executing a program stored in a memory circuit. Note that instead of storing the program in the memory circuit, the program may be directly incorporated into the circuit of the processor. In this case, the processor realizes its functions by reading and executing a program built into the circuit. Note that each processor in this embodiment is not limited to being configured as a single circuit for each processor, but may also be configured as a single processor by combining multiple independent circuits to realize its functions. good.

Further, although several embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, and changes can be made without departing from the gist of the invention. These embodiments and their modifications are included within the scope and gist of the invention as well as within the scope of the invention described in the claims and its equivalents.

1 Diagnosis support device 3 Medical image diagnosis device 5 HIS
7 RIS
9 PACS
10 I/F circuit 20 Input circuit 30 Display 40 Memory circuit 50 Processing circuit 50a Representative pattern determination function 50b Detection algorithm determination function 50c Verification processing function 50d Detection processing function 32 Rule name input area 33 Characteristic input area by text 34 Characteristic input by graph Area 35 Operation button area S Diagnosis support system

Claims (12)

A case to be detected for a subject , first diagnostic data over time regarding the subject, and at least one first characteristic indicating a type of temporal change of the first diagnostic data. An input section for inputting information,
a representative pattern determining unit that determines a first representative pattern for the case based on at least one of the first characteristics ;
Detection that determines a first detection algorithm that is an algorithm that detects a location exhibiting a predetermined temporal change form from among the first diagnostic data over time, according to the determined first representative pattern. an algorithm determination unit;
a detection unit that detects the onset of the case based on the determined first detection algorithm and the first diagnostic data over time ;
A diagnostic support device comprising: The diagnostic support device according to claim 1, wherein the representative pattern determining unit determines the first representative pattern using text information input by a user as the first characteristic . The representative pattern determining unit includes:
presenting a plurality of graph change patterns that are classified in advance and correspond to the plurality of first characteristics ;
The diagnostic support according to claim 1 or 2, wherein the first representative pattern is determined using at least one graph change pattern selected as the first characteristic from among the plurality of graph change patterns presented. Device. The representative pattern determining unit determines the first representative pattern using the outline of the graph input by the user,
The detection algorithm determining unit determines the first detection algorithm including an algorithm that calculates a similarity between the input graph outline and the graph outline of the first diagnostic data over time. 4. The diagnostic support device according to any one of 1 to 3. The input unit includes at least second diagnostic data that is temporal diagnostic data regarding the subject and is different from the first diagnostic data, and a type of temporal change of the second diagnostic data. Enter one second property and
The representative pattern determination unit determines at least a second representative pattern regarding the second diagnostic data for the case based on at least one of the second characteristics ,
The detection algorithm determining unit is configured to detect a location exhibiting a predetermined change over time from among the second diagnostic data over time in accordance with the determined second representative pattern. determining at least a detection algorithm for
A diagnostic support device according to any one of claims 1 to 4. The diagnostic support device according to claim 5, wherein the representative pattern determination unit determines the second representative pattern regarding a temporal change aspect of the second diagnostic data. 7. The method according to claim 1, further comprising a verification unit that verifies the first representative pattern and the detection algorithm using the determined first detection algorithm and a first database related to the case. The diagnostic support device according to any one of these. 8. The method according to claim 1, further comprising a verification unit that verifies the first representative pattern and the detection algorithm using the determined first detection algorithm and a second database unrelated to the case. The diagnostic support device according to any one of these. Claim 5 , further comprising a detection unit that executes a detection process of detecting data indicating the symptom or onset of the case from the first diagnostic data of the target patient using the first detection algorithm. Or the diagnostic support device according to 6 . The diagnostic support device according to claim 9, wherein the detection unit executes the detection process using at least the first detection algorithm, the second detection algorithm, and a simultaneity condition. The detection unit detects the detection when the generation time of each of the plurality of data detected using at least the first detection algorithm and the second detection algorithm is within a preset unit time. The diagnostic support device according to claim 10, which executes the process. to the computer,
Input a case to be detected for the subject, first diagnostic data over time regarding the subject, and at least one characteristic indicating a type of mode of change over time of the first diagnostic data. the step of
determining a first representative pattern for the case based on at least one of the characteristics;
Determining a first detection algorithm, which is an algorithm for detecting a portion exhibiting a predetermined temporal change form from among the first diagnostic data over time, in accordance with the determined first representative pattern. and,
detecting the onset of the case based on the determined first detection algorithm and the first diagnostic data over time;
A diagnostic support program that makes this possible.

Images