Medical information processing device, medical information processing method, and program

Abstract

To more appropriately specify the amount of a biological internal component closer to the actual body condition state of a subject.SOLUTION: A medical information processing apparatus of an embodiment has an acquisition unit, a reference data specification unit, a first derivation unit, a second derivation unit, and a reliability score determination unit. The acquisition unit acquires biological data related to a subject. The reference data specification unit specifies reference data related to a first feature quantity based on the biological data. The first derivation unit inputs the first feature quantity included in the biological data to a model that can convert the first feature quantity and a second feature quantity into each other to acquire the second feature quantity. The second derivation unit inputs the second feature quantity to the model to derive pseudo feature data related to the simulated first feature quantity. The reliability score determination unit determines a reliability score related to the biological data based on the reference data and the pseudo feature data.SELECTED DRAWING: Figure 1

Description

Technical Field

[0004] , , , , , ,

[0001] Embodiments of the present invention relate to a medical information processing apparatus, a medical information processing method, and a program.

Background Art

[0002] Conventionally, internal body components such as hemoglobin concentration and blood oxygen saturation have been important information for detecting the onset of diseases. Hemoglobin concentration can be a clue for disease diagnosis that causes skin changes derived from blood flow such as heart failure accompanied by blood flow changes and varicose veins of the lower extremities accompanied by stagnation of venous blood. In order to detect the onset of such diseases at an early stage, it is necessary to continuously monitor internal body components. In that case, a method that can be easily measured in a hospital room or at home without using special equipment is desirable. As a method for measuring the amount of internal body components without using special equipment, a method of utilizing an optical camera and a physical model to search for or determine the amount of internal skin components that minimizes the difference between the luminance value calculated by MCML (Monte Carlo modeling of light transport in multi-layered tissues) and the measured value, and a method of searching for the amount of internal skin components using a genetic algorithm are known. However, since these methods use only the difference between the calculated luminance value and the measured value as the criterion for determining the component amount, the same luminance value is obtained for a plurality of component amounts with different combinations, and the uniqueness of the solution is not satisfied, or due to the mismatch of the search range, there are cases where the amount of internal body components close to the actual physical condition of a subject such as a patient cannot be specified.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Non-Patent Documents

[0004]

Non-Patent Document 1

[0005] One of the problems that the present invention aims to solve is to more appropriately identify the amount of internal biological components that are closer to the actual physical condition of the subject. However, the problems that the embodiments disclosed in this specification and drawings aim to solve are not limited to the above problem, and problems corresponding to the effects of each configuration shown in the embodiments described later can also be positioned as other problems. [Means for solving the problem]

[0006] The medical information processing device of this embodiment includes an acquisition unit, a reference data identification unit, a first derivation unit, a second derivation unit, and a confidence score determination unit. The acquisition unit acquires biological data relating to a subject. The reference data identification unit identifies reference data relating to a first feature based on the biological data. The first derivation unit acquires a second feature by inputting the first feature contained in the biological data into a model capable of mutually converting between the first feature and the second feature. The second derivation unit derives simulated pseudo-feature data relating to the first feature by inputting the second feature into the model. The confidence score determination unit determines a confidence score relating to the biological data based on the reference data and the pseudo-feature data. [Brief explanation of the drawing]

[0007] [Figure 1] A diagram showing an example of the configuration of a medical information system 1 including a medical information processing device of an embodiment. [Figure 2] A diagram illustrating the reference data identification function 142. [Figure 3] A diagram illustrating the first derivation function 143. [Figure 4] A diagram illustrating problem analysis using a model. [Figure 5] A diagram illustrating the second derivation function 144. [Figure 6] A diagram illustrating the confidence score determination function 145. [Figure 7] This figure shows an example of the first image IM10 generated by the image generation function 146. [Figure 8] This figure shows an example of the second image IM20 generated by the image generation function 146. [Figure 9] A flowchart showing the sequence of processes executed by the processing circuit 140. [Figure 10] A diagram showing an example configuration of a medical information system 1A, including a modified medical information processing device 100A. [Figure 11] This figure shows an example of the third image IM30 generated by the image generation function 146. [Modes for carrying out the invention]

[0008] The medical information processing device, medical information processing method, and program of the embodiment will be described below with reference to the drawings.

[0009] Figure 1 shows an example of the configuration of a medical information system 1 including a medical information processing device of an embodiment. The medical information system 1 includes, for example, a patient-side terminal 10, a hospital-side terminal 20, and a medical information processing device 100. The patient-side terminal 10, the hospital-side terminal 20, and the medical information processing device 100 are connected to each other via a network NW, for example. The medical information system 1 may have multiple patient-side terminals 10 and hospital-side terminals 20.

[0010] A network (NW) refers to all information and communication networks that utilize telecommunications technology. Networks (NW) include wireless / wired LANs (Local Area Networks), WANs (Wide Area Networks), the Internet, telephone communication lines, fiber optic communication networks, cable communication networks, satellite communication networks, etc.

[0011] The patient terminal 10 acquires biometric data about the patient (an example of a subject) and transmits the acquired biometric data to the medical information processing device 100 via the network NW. The biometric data is, for example, data from which parameters of the patient's internal biological components can be acquired. The biometric data includes, for example, images of the patient's skin and information acquired by various sensors (for example, absorbance). Parameters of internal biological components are, for example, amounts of internal biological components such as hemoglobin concentration (hereinafter referred to as Hb concentration) and melanin concentration (hereinafter referred to as Me concentration). Parameters of internal biological components may also include the thickness of the patient's epidermis and dermis. Parameters of internal biological components may also include oxygen saturation in the blood and blood glucose levels.

[0012] The patient-side terminal 10 transmits biometric data in one or more time phases. The patient-side terminal 10 may also transmit patient information (e.g., identification information to identify the patient) and basic biometric data information (e.g., type of biometric data and acquisition date and time information) when transmitting biometric data. The patient-side terminal 10 is a device equipped with the function to perform the above-described processing, and may be, for example, a smartphone, tablet, camera device, or wearable device.

[0013] The hospital-side terminal 20 acquires the processing results by the medical information processing device 100 via the network NW, displays the acquired information, and provides the user such as a doctor with the patient's condition and the like. The hospital-side terminal 20 may be an installed PC (Personal Computer), a server, or the like, or may be a portable smartphone, a tablet terminal, or the like.

[0014] The medical information processing device 100 receives the biological data transmitted from the patient-side terminal 10, and performs processing such as quantifying the reliability of the internal component amounts of the living body from the biological data. Further, the medical information processing device 100 causes the processing results to be displayed on the display of its own device, or transmits them to the hospital-side terminal 20 via the network NW.

[0015] Here, the functional configuration of the medical information processing device 100 will be described. The medical information processing device 100 includes, for example, a communication interface 110, an input interface 120, a display 130, a processing circuit 140, and a memory 150.

[0016] The communication interface 110 includes, for example, a communication interface such as a NIC (Network Interface Controller). The communication interface 110 communicates with external devices such as the patient-side terminal 10 and the hospital-side terminal 20 via the network NW, and outputs the acquired information to the processing circuit 140 and the like. Further, the communication interface 110 receives control by the processing circuit 140, and transmits information to external devices such as the hospital-side terminal 20 connected via the network NW.

[0017] The input interface 120 receives various input operations from the user, converts the received input operations into electrical signals, and transmits them to the processing circuit 140. When an input operation is performed by the user, for example, the input interface 120 generates information corresponding to the input operation. The input interface 120 transmits the generated information corresponding to the input operation to the processing circuit 140. The input interface 120 is realized by, for example, a mouse, a keyboard, a trackball, a switch, a button, a joystick, a touch panel, or the like. Further, the input interface 120 may be realized by a user interface that receives voice input such as a microphone. When the input interface 120 is a touch panel, the display 130 described later may be formed integrally with the input interface 120.

[0018] The display 130 displays various information. For example, the display 130 displays an image generated by the processing circuit 140, a GUI (Graphical User Interface) for receiving various input operations from the user, and the like. For example, the display 130 is an LCD (Liquid Crystal Display), a CRT (Cathode Ray Tube) display, an organic EL (Electro Luminescence) display, or the like.

[0019] The processing circuit 140 includes, for example, an acquisition function 141, a reference data identification function 142, a first derivation function 143, a second derivation function 144, a reliability score determination function 145, an image generation function 146, and a display control function 147. The processing circuit 140 realizes these functions by, for example, a hardware processor executing a program stored in a storage device (storage circuit).

[0020] Hardware processors refer to circuits such as CPUs (Central Processing Units), GPUs (Graphics Processing Units), Application Specific Integrated Circuits (ASICs), and programmable logic devices (e.g., Simple Programmable Logic Devices (SPLDs), Complex Programmable Logic Devices (CPLDs), and Field Programmable Gate Arrays (FPGAs)).

[0021] Instead of storing the program in a memory device, the system may be configured to directly incorporate the program into the circuitry of the hardware processor. In this case, the hardware processor performs its function by reading and executing the program incorporated into the circuitry. The above program may be stored in a memory device beforehand, or it may be stored in a non-temporary storage medium such as a DVD or CD-ROM, and installed from the non-temporary storage medium to the memory device when the non-temporary storage medium is mounted in the drive device (not shown) of the medical information processing device 100.

[0022] A hardware processor is not limited to being a single circuit; it may also be composed of multiple independent circuits combined to perform various functions. Alternatively, multiple components may be integrated into a single hardware processor to perform various functions.

[0023] Memory 150 can be implemented by, for example, semiconductor memory elements such as RAM (Random Access Memory) and flash memory, hard disks, optical discs, etc. These non-transient storage media may also be implemented by other storage devices connected via a network NW, such as NAS (Network Attached Storage) or external storage server devices. Alternatively, these non-transient storage media may also be implemented by storage devices such as ROM (Read Only Memory) and registers. Memory 150 stores, for example, biometric data (Data Base) 151, model DB 152, programs, and various other types of information.

[0024] The acquisition function 141 acquires biometric data transmitted from the patient terminal 10 via the communication interface 110. The acquisition function 141 may also store the acquired biometric data in the biometric data DB 151, associating it with patient information, basic biometric data information, etc. The acquisition function 141 may also acquire biometric data from the biometric data DB 151. The biometric data DB 151 may be acquired from an external device via a network NW.

[0025] Furthermore, the acquisition function 141 may acquire information from the acquired biological data that can be transformed using a model stored in the model DB 152. The models stored in the model DB 152 are physical or mathematical models that can mutually transform the first feature and the second feature through simulation or the like. The first feature is, for example, a feature derived from the biological data. The first feature is, for example, the brightness value or absorbance of an image. The second feature is a feature derived from the first feature. The second feature is, for example, a parameter of the internal components of the biological body. Hereinafter, the acquisition of the first feature simulated by inputting the second feature into the model may be referred to as "forward problem analysis," and the acquisition of the second feature simulated by inputting the first feature into the model may be referred to as "inverse problem analysis." Also, data relating to the other feature simulated by inputting one of the first or second features into the model may be referred to as "pseudo-feature data."

[0026] The models stored in the biological data DB151 include, for example, a model that mutually converts the first and second feature quantities using a method for estimating light reflection of the skin based on Kubelkamunck theory (hereinafter referred to as the "first model"), a model that mutually converts the first and second feature quantities using light scattering simulations in biological tissue using MCML, etc. (hereinafter referred to as the "second model"), and a physical model that formulates the absorption of light by matter using the Lambert-Beer law and mutually converts the first and second feature quantities (hereinafter referred to as the "third model"). The model DB152 may be acquired from an external device via a network NW.

[0027] For example, the acquisition function 141 acquires brightness values ​​from skin images included in the biological data when the first or second model is used in subsequent processing (for example, the processing of the first derivation function 143 and the second derivation function 144), and acquires absorbance from sensor results included in the biological data when the third model is used.

[0028] Furthermore, the acquisition function 141 may perform processing (preprocessing) on ​​the biological data, such as smoothing filtering or edge extraction and removal, in order to remove information that would be noisy in deriving parameters of internal biological components contained in the biological data (for example, information about palm prints, wrinkles, and hair contained in the image). In addition, the acquisition function 141 may acquire data from a single time phase, or it may acquire data from multiple time phases in a time series.

[0029] The reference data identification function 142 identifies features (reference data) based on changes in the parameters of internal biological components. Reference data is, for example, an index value that shows a feature that reflects the change in the amount of internal biological component to be identified. Figure 2 is a diagram illustrating the reference data identification function 142. In the following, patient video (MV) will be used as biological data, with brightness value as the first feature and Hb concentration as the second feature. The brightness value is, for example, at least one brightness value among RGB. In the following, the brightness value of R (green) will be used as an example. Patient video (MV) is, for example, multiple time-phase images taken of the same area of ​​skin of a patient at different time points t (for example, skin images from which the color of blood in the skin can be extracted from the image). The patient video (MV) may include bandwidth information.

[0030] The reference data identification function 142 identifies reference data related to brightness values ​​from skin images for each time phase included in the patient video MV. For example, the blood flow in the body increases and decreases (blood flow changes) due to the cycle of contraction and expansion of the heart caused by pulsation. For example, during contraction, the blood flow increases (hemoglobin concentration increases), so the absorbance of hemoglobin in the blood increases and the brightness value decreases. On the other hand, during expansion, the blood flow decreases (hemoglobin concentration decreases), so the absorbance of hemoglobin decreases and the brightness value increases. Therefore, the brightness value of the same pixel in the image area over time also increases and decreases periodically, as shown in Figure 2. Accordingly, the reference data identification function 142 identifies information showing the change in brightness value in a predetermined area (brightness value change portion) among the time changes in brightness value caused by pulsation as reference data. This reference data may be considered as information showing the change in Hb concentration in a predetermined area.

[0031] In the example shown in Figure 2, the reference data identification function 142 extracts luminance values ​​from the time domain showing an increase in Hb concentration (the range from the maximum value (convex part) to the minimum value (concave part) of the waveform showing the increase or decrease in luminance value) among the periodic changes in luminance values. As shown in Figure 2, the reference data identification function 142 identifies the reference data by averaging the luminance values ​​of multiple time domains showing an increase in Hb concentration. This suppresses the variation in luminance values ​​across different time domains. The reference data identification function 142 may also identify the reference data by selecting any one of the multiple time domains showing an increase in Hb concentration. Alternatively, the reference data identification function 142 may identify luminance values ​​in the time domain showing a decrease in Hb concentration (the range from the minimum value to the maximum value of the luminance value) as the reference data.

[0032] The first derivation function 143 inputs luminance values ​​contained in biological data into a model to derive the amount of internal biological components, including Hb concentration. Figure 3 is a diagram illustrating the first derivation function 143. For example, the first derivation function 143 inputs the luminance values ​​of pixels at the same locations from which reference data was obtained for each time phase of patient video MV, which is biological data, into a model stored in the model DB 151 (for example, a physical model such as the first model or the second model), and derives the amount of internal biological components through problem analysis.

[0033] Figure 4 is a diagram illustrating problem analysis using a model. For example, the model used in the first derivation function 143 and the second derivation function 144 (e.g., physical model 152A) is a model that takes parameters of internal biological components (e.g., Hb concentration: A, Mel concentration: B, epidermal thickness: C, dermal thickness: D) as input and outputs luminance values ​​(RGB values) through forward problem analysis. In this case, the first derivation function 143 uses this physical model 152A to derive the internal biological component parameters from the luminance values ​​(RGB values) of pixels for each time phase through inverse problem analysis.

[0034] The second derivation function 144 takes the parameters of internal biological components derived by the first derivation function 143 as input and derives simulated brightness value data (pseudo-feature data) that can be compared with reference data. Figure 5 is a diagram illustrating the second derivation function 144. The second derivation function 144 inputs the values ​​of at least the data corresponding to the reference data (Hb concentration) from the parameters of internal biological components (for example, Hb concentration: A, Mel concentration: B, epidermal thickness: C, dermal thickness: D) at predetermined intervals into the physical model 152A, and derives the values ​​obtained repeatedly by forward problem analysis of the physical model 152A as pseudo-feature data (referred to as "patient simulation data" in Figure 5) for the simulated first feature. The predetermined amount may be a fixed amount or may be set variably for each internal biological component. The number of times the value is changed at predetermined intervals (number of repetitions) may be set based on the time course of the reference data to be compared, or it may be a fixed number of repetitions.

[0035] The second derivation function 144 may generate a set of internal biological component amounts by increasing the Hb concentration by a predetermined amount for the number of times mentioned above, input the generated set into the model, and derive luminance values ​​for the set of internal biological component amounts. Alternatively, the second derivation function 144 may normalize the range of luminance values ​​(0 to 255) to a predetermined range (for example, the range of 0 to 1), as shown in the example in Figure 5. In the example in Figure 5, the second derivation function 144 increases the Hb concentration by a predetermined increase amount ΔA and derives the changes in each simulated luminance value () as pseudo-feature data by inputting it into the physical model 152A. In the example in Figure 5, since the reference data was the change from the maximum to the minimum of the luminance change period due to pulsation, the pseudo-feature data is similarly derived from the maximum to the minimum. However, if the reference data is in the range from the minimum to the maximum, the second derivation function 144 derives simulated pseudo-feature data by gradually decreasing the Hb concentration by a predetermined amount ΔA.

[0036] The confidence score determination function 145 determines the confidence score of the amount of internal biological components from the reference data derived by the first derivation function 143 and the pseudo-feature data derived by the second derivation function 144. Figure 6 is a diagram illustrating the confidence score determination function 145. The confidence score determination function 145 compares the reference data with the pseudo-feature data (patient data simulation values), evaluates how similar the two input data are (how reliable the Hb concentration is), and determines an index value (e.g., confidence score) that quantifies the similarity of each data. For example, the confidence score determination function 145 may determine the confidence score from the dot product of the two input values. In this case, the confidence score determination function 145 assigns a higher confidence score the closer the dot product is to 1.

[0037] Furthermore, the confidence score determination function 145 may determine the confidence score using the reciprocal of the mean absolute error (MAE) of the two input values ​​(1 / MAE), or it may determine the confidence score using the reciprocal of the mean squared error (MSE) (1 / MSE). In this case, the confidence score determination function 145 will assign a higher confidence score the larger the value of the reciprocal. Also, for each change in input value, the confidence score determination function 145 will assign a higher confidence score the closer the sum of the slopes between two adjacent points is. In addition, the confidence score determination function 145 may use other methods to calculate the degree of similarity between the two input values.

[0038] The confidence score determination function 145 can obtain a confidence score for the entire image by performing the above-described process for each pixel of the image included in the patient video MV. The confidence score determination function 145 may also determine the score only for a predetermined image range from the entire image for which a confidence score is required. This image range may be determined based on the imaging position of the patient video MV, or it may be set by the user via the input interface 120.

[0039] The image generation function 146 generates an image containing information about the confidence score determined by the confidence score determination function 145. Figure 7 shows an example of the first image IM10 generated by the image generation function 146. The content, layout, color, design, and other display characteristics of the image IM10 described below are not limited to this example. The same applies to other images described later.

[0040] Image IM10 shown in Figure 7 includes, for example, a patient information display area AR11, a biometric data display area AR12, a confidence score display area AR13, a biometric internal component quantity display area AR14, a setting input area AR15, and a processing result display area AR16. The patient information display area AR11 displays identification information that identifies the patient from whom biometric data was acquired (e.g., patient ID) and the date and time the biometric data was acquired (e.g., the date and time the skin image was taken). The biometric data display area AR12 displays the target image (skin image IM11 in the example in Figure 7) from which the confidence score included in the biometric data DB151 was determined.

[0041] The confidence score display area AR13 displays the distribution image IM12 of the confidence scores for the skin image IM11. The internal biological component amount display area AR14 displays images IM14 and IM15 showing the derivation results of internal biological component amounts (e.g., Hb concentration) for the skin image IM11. The setting input area AR15 displays an image for the user to set the lower limit of the confidence score. In the example in Figure 7, a slider is displayed that allows the user to adjust the lower limit of confidence by inputting operations into the input interface 120 or by instructions from the hospital terminal 20. For example, the image generation function 146 displays image IM14, which is an image showing the Hb concentration derivation results, with the area where the confidence score is less than the lower limit of confidence masked, based on the lower limit of confidence set by the user. Masking can be done by, for example, overlaying other images, removing them, or hiding them. This allows the user to grasp the Hb concentration with high confidence. The processing result display area AR16 displays processing results such as Hb density and brightness values ​​for image areas above the confidence threshold.

[0042] The display control function 147 displays the image IM10 generated by the image generation function 146 on the display 130 or transmits it to the hospital terminal 20 via the network NW. The display control function 147 may also store processing results etc. in the memory 150, and may display the information stored in the memory 150 on the display 130 or transmit it to the hospital terminal 20.

[0043] By displaying the first image IM10, for example, the amount of internal components, including confidence scores, can be provided to physicians and other medical professionals. As a result, physicians can exclude low-confidence data from the amount of internal components and make diagnoses based on high-confidence data, thus preventing missed diagnoses of disease onset in patients and enabling more accurate diagnoses.

[0044] The image generation function 146 may generate other images in place of (or in addition to) the first image IM10. Figure 8 shows an example of the second image IM20 generated by the image generation function 146. The second image IM20 contains information prompting the patient to reacquire vital data (in the example in Figure 8, to retake the skin image).

[0045] The second image IM20 shown in Figure 8 includes, for example, a patient information display area AR21, a biometric data display area AR22, a confidence score display area AR23, a biological component amount display area AR24, and a processing result display area AR25. The patient information display area AR21, the biometric data display area AR22, and the confidence score display area AR23 display the same information as the display areas AR11 to AR13 of the first image IM10 described above. The biological component amount display area AR24 displays image IM14, which shows the result of deriving the biological component amount (e.g., Hb concentration) for the skin image IM11. The processing result display area AR25 displays confidence score information (numerical value) for the biological component amount of the processed image, and information on whether rescanning is necessary based on the processing result.

[0046] For example, if the confidence score for the amount of internal biological components in the skin image IM11 is below a threshold, the image generation function 146 displays information prompting the patient to reacquire the biological data (skin image). In the example in Figure 8, since the confidence score (34.5 ± 13 (%)) is below the threshold, information indicating that reacquisition of the biological data is necessary ("Warning: There may be an image capture error. Reacquisition is required.") is displayed in the internal biological component amount display area AR24.

[0047] The display control function 147 may, when a second image IM20 is generated, display it on the display 130, send it to the hospital terminal 20, or send it to the patient terminal 10 to prompt the patient to reacquire and retransmit the biometric data. By displaying the second image IM20, the patient is prompted to reacquire the input data if the confidence score is low, thereby preventing misdiagnosis due to information with a low confidence score, or missing the onset of disease (for example, the onset of heart failure or the onset of varicose veins in the lower extremities).

[0048] [Processing flow] The following describes the processing flow of the processing circuit 140 in the embodiment. Figure 9 is a flowchart showing a series of processes performed by the processing circuit 140. In the example in Figure 9, the acquisition function 141 acquires a skin image (a moving image including images at multiple time phases) in order to obtain the amount of internal biological components from the biological data DB 151 (step S100). Next, the reference data identification function 142 identifies the portion of the acquired skin image where the brightness value changes due to the increase in Hb concentration due to pulsation (reference data) (step S110).

[0049] Next, the first derivation function 143 selects a target pixel identical to the pixel from which the reference data was acquired (step S120), and derives the Hb concentration for the brightness value of the selected target pixel (step S130). The first derivation function 143 also generates a set of biological internal component amounts with increased Hb concentration (step S140).

[0050] Next, the second derivation function 144 derives brightness values ​​for the set of internal biological component amounts and calculates patient data simulation values ​​(pseudo-feature data) (step S150). Next, the confidence score determination function 145 determines a confidence score based on the reference data and the patient data simulation values ​​(step S160). Next, the image generation function 146 generates an image including the confidence score (step S170). Next, the display control function 147 displays the generated image on the display 130 (step S180). In step S180, the generated image may be transmitted to an external device (hospital-side terminal 20) via the network. This completes the processing of this flowchart.

[0051] [Differentiation] The medical information processing device 100 of this embodiment may, for example, use the confidence score for biological data calculated by the confidence score determination function 145 when training a model for CDS (Clinical Decision Support). The above content will be described below as a modified version of the medical information processing device. Configurations similar to those described in the medical information system 1 above will be given the same names and reference numerals, and specific explanations will be omitted here.

[0052] Figure 10 shows an example configuration of a medical information system 1A including a modified medical information processing device 100A. The medical information system 1 includes, for example, a patient-side terminal 10, a hospital-side terminal 20, and a medical information processing device 100A. The patient-side terminal 10, the hospital-side terminal 20, and the medical information processing device 100A are communicated together, for example, via a network NW.

[0053] The medical information processing device 100A includes, for example, a communication interface 110, an input interface 120, a display 130, a processing circuit 140A, and a memory 150. The processing circuit 140A includes, for example, an acquisition function 141, a reference data identification function 142, a first derivation function 143, a second derivation function 144, a confidence score determination function 145, an image generation function 146, a display control function 147, and a learning function 148. The processing circuit 140A differs from the processing circuit 140 of the medical information processing device 100 in that it has a learning function 148. Therefore, the following explanation will mainly focus on the learning function 148.

[0054] The learning function 148 trains a model using biometric data adjusted based on the confidence score determined by the confidence score determination function 145. For example, the learning function 148 trains a CDS model or other models (e.g., models stored in the model DB 152) using the confidence score determined by the confidence score determination function 145. The CDS model is used in settings that perform image analysis, such as medical facilities (e.g., hospital terminals 20), beauty salons, and other computer-aided diagnostic (CAD) systems.

[0055] The learning function 148 acquires skin images and performs preprocessing such as smoothing filtering and edge detection / removal. The learning function 148 also splits the data into training and test data, and selects regions to be used for training based on the confidence score determined by the confidence score determination function 145. Specifically, it masks low-confidence regions from the skin images to be trained where the confidence score of the amount of internal biological components is below a threshold (e.g., 75%). Then, the learning function 148 trains the CDS model using the unmasked data with confidence scores above the threshold. For training, well-known learning methods such as deep learning or other machine learning techniques can be used. The learning function 148 may also evaluate the accuracy of the training results.

[0056] The image generation function 146 generates an image that includes the results processed by the learning function 148. Figure 11 shows an example of a third image IM30 generated by the image generation function 146. In the example in Figure 11, image IM30 includes a learning content display area AR31 and a learning result display area AR32. The learning content display area AR31 displays the flow of the learning process performed by the learning function 148 and the results of each process. In the example in Figure 11, the processing results show a skin image IM11 and an image IM14 in which areas with low confidence scores are masked.

[0057] The learning result display area AR32 displays, for example, the evaluation result (accuracy) of the trained model based on the number of training data samples and whether or not screening (masking based on confidence scores) was performed. The display control function 147 displays the generated image on the display 130 or transmits it to an external device via the network NW.

[0058] According to the modified example described above, by providing the user with information such as image IM30, the confidence score can be used to determine whether or not to use it for model training. The medical information processing device 100A may also accept a selection from the user via the input interface 120 whether to use screened data or unscreened data for model training, and execute a training process corresponding to the accepted result.

[0059] The medical information processing devices 100 and 100A of this embodiment may include at least some of the functions of the patient-side terminal 10 and at least some of the functions of the hospital-side terminal 20. Therefore, the medical information processing devices 100 and 100A may be provided with a function to capture the patient's biological data (e.g., patient video images, etc.), and doctors and others may perform diagnoses of patients while viewing the images displayed on the display 130 of the medical information processing devices 100 and 100A.

[0060] Furthermore, although the above-described embodiment mainly used Hb concentration as a parameter for internal biological components, instead (or in addition to) this, confidence scores may be determined for other amounts of internal biological components that can be obtained as biological data (for example, Mel concentration, blood glucose level, venous blood oxygen saturation (SpO2), and arterial blood oxygen saturation (SvO2). In this case, the medical information processing devices 100 and 100A of the embodiment change the parameters of the internal biological components for which the confidence score is calculated according to the content of the biological data obtained by the acquisition function 141 and the content of the reference data identified by the reference data identification function 142. For example, if the reference data is absorbance detected by a sensor, then oxygen saturation in the blood (for example, venous blood oxygen saturation, arterial blood oxygen saturation) is selected as the parameter for internal biological components. In addition, the medical information processing devices 100 and 100A of the embodiment may receive information from the user via the input interface 120 regarding the subject for which the confidence score is to be determined, and perform processing using the information corresponding to the received subject.

[0061] Furthermore, in the above-described embodiment, reference data was identified based on the assumption of a change in brightness value due to pulsation. However, instead, reference data may be identified based on the assumption of a change in brightness value due to compression of a blood vessel using a cuff or other component (an increase in brightness value due to a decrease in blood flow).

[0062] Furthermore, the medical information processing devices 100 and 100A of the embodiment may determine a confidence score based on reference data identified from a single-temporal image and pseudo-feature data derived from the same image. Alternatively, the medical information processing devices 100 and 100A of the embodiment may determine a confidence score based on multiple reference data and multiple pseudo-feature data. In this case, the reference data identification function 142 identifies reference data for each of multiple first features based on the biological data. The first derivation function 143 derives multiple second features by inputting each of the first features into a model, and the second derivation function 144 derives multiple pseudo-feature data by inputting the multiple second features into a model. Finally, the confidence score determination function 145 determines a confidence score based on the multiple reference data and multiple pseudo-feature data. In this way, comparisons can be made under various conditions using a single or multiple temporal phases, allowing for the determination of a more detailed confidence score and enabling more appropriate identification of the amount of internal biological components that are closer to the actual physical condition of the subject.

[0063] In the embodiments described above, the acquisition function 141 is an example of an "acquisition unit," the reference data identification function 142 is an example of a "reference data identification unit," the first derivation function 143 is an example of a "first derivation unit," the second derivation function 144 is an example of a "second derivation unit," the confidence score determination function 145 is an example of a "confidence score determination unit," the image generation function 146 is an example of an "image generation unit," the display control function 147 is an example of a "display control unit," and the learning function 148 is an example of a "learning unit."

[0064] According to at least one embodiment described above, the medical information processing device of the embodiment includes: an acquisition unit that acquires biological data relating to a subject; a reference data identification unit that identifies reference data relating to a first feature based on the biological data; a first derivation unit that acquires a second feature by inputting the first feature included in the biological data into a model capable of mutually converting between the first feature and the second feature; a second derivation unit that derives simulated pseudo-feature data relating to the first feature by inputting the second feature into the model; and a confidence score determination unit that determines a confidence score relating to the biological data based on the reference data and the pseudo-feature data. By having these components, it is possible to more appropriately identify the amount of internal biological components that are closer to the actual physical condition of the subject.

[0065] Specifically, according to the embodiment, features based on changes in the amount of internal biological components are derived as reference data, and simulated pseudo-feature data regarding the amount of internal biological components is derived using a model. By quantifying the reliability of the amount of internal biological components from each data set, low-reliability amounts of internal biological components can be excluded. This prevents doctors from overlooking the onset of diseases in patients during their diagnoses, thus enabling more accurate diagnoses.

[0066] Furthermore, according to the embodiment, by providing information that prompts the reacquisition of biometric data with a low confidence score, appropriate biometric data can be obtained, and the amount of internal biological components that is closer to the subject's actual physical condition can be identified more accurately. Therefore, it is possible to prevent overlooking the onset of disease in patients.

[0067] Furthermore, according to this embodiment, by using only information with high confidence scores for model training, a more accurate model can be obtained.

[0068] The embodiments described above can be expressed as follows. Memory to store the program, Equipped with a processor, The processor executes the program, Obtain biometric data about the subject, Based on the aforementioned biometric data, reference data for the first feature is identified. The first feature contained in the aforementioned biological data is input into a model capable of mutually converting between the first feature and the second feature to obtain the second feature. By inputting the aforementioned second feature into the model, pseudo-feature data relating to the simulated first feature is derived. A confidence score for the biometric data is determined based on the aforementioned reference data and the aforementioned pseudo-feature data. Medical information processing device.

[0069] While several embodiments have been described, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These embodiments can be carried out in a variety of other forms, and various reductions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims and their equivalents. [Explanation of Symbols]

[0070] 1, 1A…Medical information system, 10…Patient terminal, 20…Hospital terminal, 100, 100A…Medical information processing device, 110…Communication interface, 120…Input interface, 130…Display, 140, 140A…Processing circuit, 141…Acquisition function, 142…Reference data identification function, 143…First derivation function, 144…Second derivation function, 145…Confidence score determination function, 146…Image generation function, 147…Display control function, 148…Learning function, 150…Memory

Claims

1. An acquisition unit that acquires biological data about the subject, A reference data identification unit identifies reference data for a first feature based on the aforementioned biological data, A first derivation unit obtains a second feature by inputting the first feature contained in the biological data into a model capable of mutually converting between the first feature and the second feature, A second derivation unit that derives pseudo-feature data relating to the first feature simulated by inputting the second feature into the model, A confidence score determination unit that determines a confidence score for the biological data based on the reference data and the pseudo-feature data, A medical information processing device equipped with [a specific feature].

2. The aforementioned biological data is single or multiple time-series biological data. The medical information processing device according to claim 1.

3. The aforementioned first feature is a feature derived from the biological data, The two features mentioned above are features derived from the first feature mentioned above. The medical information processing device according to claim 1.

4. The aforementioned reference data identification unit identifies reference data for each of the multiple first feature quantities based on the biological data, The first derivation unit derives a plurality of second features by inputting each of the first features into the model, The second derivation unit derives a plurality of pseudo-feature data by inputting a plurality of the second feature quantities into the model, The confidence score determination unit determines the confidence score based on a plurality of reference data and a plurality of pseudo-feature data. The medical information processing device according to claim 1.

5. The aforementioned biological data includes images, The confidence score determination unit determines a confidence score for each pixel included in the image, or for a predetermined image range. The medical information processing device according to claim 1.

6. The system further includes an image generation unit that generates an image containing information about the confidence score determined by the confidence score determination unit. The medical information processing device according to claim 1.

7. The image generation unit generates an image in which the region of the image acquired as the biological data of the subject that has the confidence score below the threshold is masked. The medical information processing device according to claim 6.

8. The image generation unit generates an image that prompts the reacquisition of the biometric data when the confidence score determined by the confidence score determination unit is below a threshold. The medical information processing device according to claim 6.

9. The system further includes a learning unit that trains a model using biometric data adjusted based on the confidence score determined by the confidence score determination unit. The medical information processing device according to claim 1.

10. Computers Obtain biometric data about the subject, Based on the aforementioned biometric data, reference data for the first feature is identified. The first feature contained in the aforementioned biological data is input into a model capable of mutually converting between the first feature and the second feature to obtain the second feature. By inputting the second feature into the model, pseudo-feature data relating to the first feature is derived. A confidence score for the biometric data is determined based on the aforementioned reference data and the aforementioned pseudo-feature data. Medical information processing method.

11. On the computer, Obtain biometric data about the subject, Based on the aforementioned biological data, reference data for the first feature is identified. The first feature contained in the aforementioned biological data is input into a model capable of mutually converting between the first feature and the second feature to obtain the second feature. By inputting the second feature into the model, pseudo-feature data relating to the simulated first feature is derived. A confidence score for the biometric data is determined based on the aforementioned reference data and the aforementioned pseudo-feature data. program.

Images