Decision prediction system, decision prediction method, and decision prediction program

Abstract

To provide a determination prediction system, determination prediction method, determination prediction program for supporting determination on diseases.SOLUTION: In a determination prediction system, a control section 21 of a support server performs: generating a plurality of division images obtained by dividing an original image; generating division prediction models for predicting a determination result from the division images with the use of first teacher information recorded by association with a determination result in the original image with respect to the plurality of division images; recording the models in a learning result storage section 24; generating second teacher information obtained by summarizing feature information which is generated by inputting the respective division images of the first teacher information to the division prediction models, based on the positions of the division images; generating an integrated prediction model for predicting a determination result with the use of the second teacher information; and recording the model in the learning result storage section 24. When an evaluation object image is acquired, the control section 21 predicts a determination result with the use of the division prediction models and the integrated prediction model recorded in the learning result storage section 24.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a diagnosis prediction system, a diagnosis prediction method, and a diagnosis prediction program for assisting in diagnosis of a disease. [Background technology]

[0002] In recent years, technology for judging images using a computer has been considered. Recently, in the medical field, support for diagnosis using images has been promoted, and information processing devices that present areas that contributed to the judgment along with the judgment result regarding the diagnosis of a disease are also being considered (for example, see Patent Document 1). The information processing device described in this document inputs an endoscopic image into a first model that outputs a diagnostic criterion prediction regarding the diagnostic criterion of a disease, and outputs the diagnostic criterion prediction. In this case, the area that influenced the diagnostic criterion prediction is extracted, and the diagnostic criterion prediction, an index indicating the extracted area, and a diagnostic prediction regarding the state of the disease are output in association with each other. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2020-89712 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, depending on the image, it may be difficult to detect diseases with high accuracy. For example, traditionally, doctors interpret chest X-ray images to determine the possibility of disease. X-ray images are two-dimensional images that contain less information than 3D CT images, and it may be difficult to make a judgment using image processing techniques. [Means for solving the problem]

[0005] The judgment prediction system for solving the above problem includes a learning result storage unit that records a division prediction model and an integrated prediction model, and a control unit that evaluates an image. The control unit generates a plurality of divided images by dividing an original image, generates first teacher information that associates the judgment results of the original image with the plurality of divided images and records it, generates a division prediction model that predicts the judgment result from the divided images using the first teacher information, and records it in the learning result storage unit, generates second teacher information that summarizes feature information generated by inputting each divided image of the first teacher information into the division prediction model based on the position of the divided image, generates an integrated prediction model that predicts the judgment result using the second teacher information, and records it in the learning result storage unit, and when an evaluation target image is acquired, predicts the judgment result using the division prediction model and the integrated prediction model recorded in the learning result storage unit. Effect of the Invention

[0006] According to the present invention, it is possible to assist in making judgments regarding diseases. [Brief description of the drawings]

[0007] [Figure 1] FIG. 1 is an explanatory diagram of a judgment prediction system according to an embodiment. [Diagram 2] FIG. 2 is an explanatory diagram of a hardware configuration of the embodiment. [Diagram 3] FIG. 4 is an explanatory diagram of image division according to the embodiment. [Figure 4] FIG. 4 is an explanatory diagram of a learning process according to the embodiment. [Diagram 5] FIG. 4 is an explanatory diagram of a procedure of a division learning process according to the embodiment. [Figure 6] FIG. 4 is an explanatory diagram of a division prediction model according to the embodiment. [Figure 7] FIG. 4 is a diagram illustrating an integrated learning process according to the embodiment. [Figure 8] FIG. 4 is an explanatory diagram of creating a heat map according to the embodiment. [Figure 9] FIG. 4 is an explanatory diagram of a heat map according to the embodiment. [Figure 10] FIG. 4 is an explanatory diagram of an integrated prediction model according to the embodiment. [Figure 11] FIG. 4 is an explanatory diagram of a prediction process according to the embodiment. [Figure 12] FIG. 11 is an explanatory diagram of a processing procedure according to another embodiment. [Figure 13] FIG. 11 is an explanatory diagram of a processing procedure according to another embodiment. [Figure 14] FIG. 11 is an explanatory diagram of a processing procedure according to another embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] An embodiment of a diagnosis prediction system, a diagnosis prediction method, and a diagnosis prediction program will be described with reference to Figures 1 to 11. In this embodiment, training data is used to associate X-ray images (organ images) from a lung cancer screening with the results of the screening, and a prediction model for predicting the possibility of lung cancer is generated. Then, when a new X-ray image is acquired in the lung cancer screening, the prediction model is used to predict the diagnosis result. As shown in FIG. 1, a user terminal 10 and a support server 20 are used, which are connected via a network.

[0009] (Hardware configuration example) FIG. 2 shows an example of the hardware configuration of an information processing device H10 that functions as the user terminal 10, the support server 20, and the like.

[0010] The information processing device H10 includes a communication device H11, an input device H12, a display device H13, a storage device H14, and a processor H 15. Note that this hardware configuration is an example, and the information processing device H10 may include other hardware.

[0011] The communication device H11 is an interface that establishes a communication path with other devices and transmits and receives data, and is, for example, a network interface card or a wireless interface.

[0012] The input device H12 is a device that accepts input from a user, etc., and is, for example, a mouse, a keyboard, etc. The display device H13 is, for example, a display or a touch panel that displays various information.

[0013] The storage device H14 is a storage device (e.g., a teacher information storage unit 23 and a learning result storage unit 24 described later) that stores data and various programs for executing various functions of the user terminal 10 and the support server 20. Examples of the storage device H14 include a ROM, a RAM, a hard disk, etc.

[0014] The processor H15 uses programs and data stored in the storage device H14 to control each process (for example, a process in the control unit 21 described later) in the user terminal 10 and the support server 20. Examples of the processor H15 include a CPU and an MPU. The processor H15 loads a program stored in a ROM or the like into a RAM and executes various processes corresponding to the various processes. For example, when an application program in the user terminal 10 or the support server 20 is started, the processor H15 operates a process that executes each process described later.

[0015] The processor H15 is not limited to a processor that performs software processing for all the processes it executes. For example, the processor H15 may be equipped with a dedicated hardware circuit (e.g., an application specific integrated circuit: ASIC) that performs hardware processing for at least a part of the processes it executes. That is, the processor H15 may be configured as a circuit including (1) one or more processors that operate according to a computer program (software), (2) one or more dedicated hardware circuits that execute at least a part of various processes, or (3) a combination thereof. The processor includes a CPU and memory such as RAM and ROM, and the memory stores program code or instructions configured to cause the CPU to execute processes. The memory, i.e., computer-readable medium, includes any available medium that can be accessed by a general-purpose or dedicated computer.

[0016] (Functions of each information processing device) The user terminal 10 in FIG. 1 is a computer terminal used by a person in charge of lung cancer screening. The support server 20 is a computer system for supporting lung cancer screening. The support server 20 includes a control unit 21, an image information storage unit 22, a teacher information storage unit 23, a learning result storage unit 24, and an evaluation target storage unit 25.

[0017] The control unit 21 performs processes (including an acquisition stage, a learning stage, a prediction stage, etc.) described below. By executing a determination prediction program for this purpose, the control unit 21 functions as an acquisition unit 211, a learning unit 212, a prediction unit 213, etc.

[0018] The acquisition unit 211 executes a process of acquiring X-ray images of learning objects and evaluation objects. The acquisition unit 211 divides an X-ray image including the entire lung field (whole lung field image as an original image) into four lung field area images (divided lung field images) from the upper left to the lower right by pattern recognition. In this pattern recognition, anatomical feature points of the lung field area are detected from the X-ray image, and the lung field is divided according to the lung field area based on the feature points.

[0019] The learning unit 212 executes a process of generating a prediction model for predicting the determination results (negative probability and positive probability) of a lung cancer screening using the teacher data. The prediction unit 213 executes a process of predicting the judgment result of the interpretation of the X-ray image to be evaluated, using the prediction model.

[0020] X-ray images taken during lung cancer screening are recorded in the image information storage unit 22. The X-ray images are associated with an examinee identifier and a diagnosis result. The examinee identifier is an identifier for identifying the examinee for the lung cancer screening who obtained this X-ray image.

[0021] The result of the determination is recorded as diagnostic information by the radiologist. This diagnostic information includes (a) the presence or absence of abnormal findings, (b) the location of the abnormal shadow, and (c) the details of the abnormal findings. Here, for (a) the presence or absence of abnormal findings, determination results such as "no abnormal findings", "suspected lung cancer", "abnormal findings are found, and a condition that requires treatment for a disease other than lung cancer is considered" are included. Note that in this embodiment, a case where the presence or absence of "suspected lung cancer" (positive or negative) is used will be described. Also, for (b) the location of the abnormal shadow, a case where the lung field region corresponding to the location of the abnormal shadow (upper right, upper left, lower right, lower left) is used when a positive determination is made will be described.

[0022] The teacher information storage unit 23 records teacher data used in a learning process for generating a prediction model for predicting the determination results (negative probability and positive probability) of a lung cancer screening. In this embodiment, the teacher data for divided learning and the teacher data for integrated learning are recorded.

[0023] The teacher data for division learning is used when generating a division prediction model for making a positive determination for each lung field region (upper right lung field, upper left lung field, lower right lung field, and lower left lung field). The teacher data for integrated learning is used when generating an integrated prediction model that integrates and judges the prediction results from the divided prediction models for each lung field.

[0024] As shown in Fig. 3, an original X-ray image 500 is divided into divided lung field images 510 (upper right lung field image 511, upper left lung field image 512, lower right lung field image 513, and lower left lung field image 514). Furthermore, a positive flag is associated with the divided lung field image (here, lower left lung field image 514) depending on the position of the abnormal shadow. Meanwhile, a negative flag is associated with the other divided lung field images. The resulting data is then recorded in the teacher information storage unit 23 as teacher data for division learning.

[0025] The learning result storage unit 24 records prediction models (division prediction models for each lung field and integrated prediction model) generated by machine learning using teacher data. The X-ray image of the evaluation target and the prediction result are recorded in the evaluation target storage unit 25. The X-ray image of the evaluation target is recorded when it is newly acquired from the user terminal 10, and the prediction result is recorded at the prediction stage described later.

[0026] Next, a process for supporting a diagnosis in a lung cancer screening in the system configured as above will be described.

[0027] (Learning process) The learning process will be described with reference to FIG. First, the control unit 21 of the support server 20 executes an image acquisition process (step S101). Specifically, the acquisition unit 211 of the control unit 21 acquires an X-ray image of a lung cancer screening from the user terminal 10. This X-ray image (whole lung field image) includes an examiner identifier (anonymous) and a positive determination result by an image interpretation doctor. If a positive determination is made, information for identifying the determined lung field is included. The acquisition unit 211 records the acquired whole lung field image in the image information storage unit 22 in association with the examiner identifier and the determination result.

[0028] Next, the control unit 21 of the support server 20 executes a segmentation process for each entire lung field image recorded in the image information storage unit 22 (step S102). Specifically, the acquisition unit 211 of the control unit 21 segments the entire lung field image into an upper left lung field image, a lower left lung field image, an upper right lung field image, and a lower right lung field image according to the lung field region by pattern recognition. The acquisition unit 211 then temporarily stores the upper left to lower right segmented lung field images in memory in association with the examinee identifier. In this case, a positive flag is recorded in the segmented lung field images according to the position of the abnormal shadow in the entire lung field image. Also, a negative flag is recorded in the segmented lung field images of the other lung fields. The above process is repeated until it is completed for all the entire lung field images.

[0029] Next, the control unit 21 of the support server 20 executes a process of generating first teacher information for each lung field region (step S103). Specifically, the acquisition unit 211 of the control unit 21 compiles the divided lung field images temporarily stored in the memory for each lung field region to generate teacher data for division learning (first teacher information). Then, the acquisition unit 211 records the generated teacher data for division learning in the teacher information storage unit 23 in association with the divided lung fields from the upper left to the lower right.

[0030] Next, the control unit 21 of the support server 20 executes a segmentation learning process for each of the segmented lung fields from the upper left to the lower right (step S104). This process will be described later with reference to FIG. Next, the control unit 21 of the support server 20 executes an integrated learning process (step S105). This process will be described later with reference to FIG.

[0031] (Divided learning process) The division learning process will be described with reference to FIG. First, the control unit 21 of the support server 20 executes a process of acquiring the first teacher information (step S201). Specifically, the learning unit 212 of the control unit 21 acquires teacher data for division learning recorded in the teacher information storage unit 23.

[0032] Next, the control unit 21 of the support server 20 executes a process of generating a feature map image in multiple layers (step S202). Specifically, the learning unit 212 of the control unit 21 uses "DenseNet" as a machine learning method. This "DenseNet" is composed of a network in which "Dense Block" and "Transition Layer" are alternately sandwiched between a convolution layer and a pooling layer. Using "DenseNet", a prediction model ("DenseNet" model) is generated in which a segmented lung field image is input as an input layer and a positive / negative is output in an output layer. In this case, the "DenseNet" model to which the segmented image is input generates a feature map image in an intermediate layer consisting of multiple layers according to the depth. The feature map image is an image in which various filters are applied to the segmented lung field image for each layer. Furthermore, in the "DenseNet" model, a representative value of the feature map image for each layer is calculated. Then, in the "DenseNet" model, a weighted value (variable value within the model) for each layer that outputs a positive / negative probability is calculated from this representative value.

[0033] Next, the control unit 21 of the support server 20 executes a process of acquiring a representative value for each hierarchical level of the feature map image (step S203). Specifically, the learning unit 212 of the control unit 21 acquires a representative value for each feature map image for each hierarchical level of the "DenseNet" model. In this embodiment, the maximum image value in the feature map image is used as the representative value.

[0034] Next, the control unit 21 of the support server 20 executes a process of acquiring weights for each hierarchical level, which uses the representative value for each hierarchical level as an input and outputs the positive / negative probability (step S204). Specifically, the learning unit 212 of the control unit 21 acquires weights (weighting information) for calculating the likelihood (probability) of positive and negative using the representative value of the feature map image for each hierarchical level in the "DenseNet" model. The feature map image that influences the positive and negative probabilities is weighted by the weights.

[0035] As shown in FIG. 6, the learning unit 212 calculates weights (W01-W04) that output a negative probability and weights (W11-W14) that output a positive probability in response to input of representative values ​​(vd1-vd4) of the feature map image 520 of the divided lung image 510. In this case, each divided lung image 510 is input to the "DenseNet" model to generate a feature map image of each layer, and a representative value of this feature map image is calculated. Furthermore, the positive probability and negative probability output by the "DenseNet" model are obtained. Next, the learning unit 212 obtains the weights (W01-W14) from the "DenseNet" model. Note that, although four layers are assumed in FIG. 6, when the "DenseNet" model has n layers, the number of representative values ​​becomes n and the number of weights becomes 2n.

[0036] Then, the learning unit 212 records in the learning result storage unit 24 a network (segmentation prediction model) including the “DenseNet” model and weighting values ​​for the segmented lung field identifiers and hierarchical identifiers.

[0037] (Integrated learning process) The integrated learning process will be described with reference to FIG. Here, the control unit 21 of the support server 20 executes a process of creating a heat map for each lung field image of each examinee using a feature map image and a weighting value (step S301). Specifically, the learning unit 212 of the control unit 21 creates a heat map for each examinee according to the lung field region using the feature map image of each layer generated by inputting each divided lung field image into the "DenseNet" model and the weighting value for calculating the positive probability in this layer.

[0038] 8, a weighted image is generated by multiplying each feature amount map image 520 of each layer by a weight value (W11 to W14) that calculates a positive probability. Furthermore, the learning unit 212 generates a heat map 600 by overlaying the weighted images of each layer at the same pixel position.

[0039] 9 shows an example of a heat map 600 in which the divided lung field images of each examinee are combined with the original whole lung field image 500. In this heat map 600, multiple feature maps are superimposed, and the feature regions of the feature map images to which high weighting values ​​are assigned are displayed in an emphasized manner.

[0040] Next, the control unit 21 of the support server 20 executes a registration process of the second teacher information (for integrated learning) (step S302). Specifically, the learning unit 212 of the control unit 21 generates teacher data for integrated learning (second teacher information) using the division learning result. Here, the heat map and the judgment result for each lung field region compiled for each examinee created in step S301 are used as the teacher data for integrated learning. Then, the learning unit 212 records the heat map and the judgment result for each lung field region associated with the examinee identifier in the teacher information storage unit 23 as the teacher data for integrated learning. The above process is repeated for all examinees recorded in the image information storage unit 22.

[0041] Next, the control unit 21 of the support server 20 executes a process of creating an integrated prediction model (step S303). Specifically, the learning unit 212 of the control unit 21 generates an integrated prediction model that predicts a negative probability and a positive probability by machine learning using teacher data for integrated learning generated using the result of the divided learning.

[0042] Here, as shown in Fig. 10, the feature amount (integrated input feature amounts vi1 to vi4) for each lung field region of each feature amount map image 520 is calculated, and an integrated prediction model that outputs a negative probability and a positive probability using each integrated input feature amount as an input is generated by machine learning. In this embodiment, machine learning is performed using the heat map for each lung field region created in step S301 as the integrated input feature amounts vi1 to vi4 as an input, and the judgment result of the radiologist as an output. Note that in Fig. 10, the lung field image is divided into four parts from the upper left to the lower right, so there are four feature amounts, but the number of feature amounts is determined according to the number of divisions of the lung field. Then, the learning unit 212 records the generated integrated prediction model in the learning result storage unit 24.

[0043] (Prediction processing) Next, the prediction process will be described with reference to FIG.

[0044] First, the control unit 21 of the support server 20 executes an image acquisition process (step S401). Specifically, the acquisition unit 211 of the control unit 21 acquires lung field images in a new lung cancer screening from the user terminal 10, and records them in the evaluation target storage unit 25.

[0045] Next, the control unit 21 of the support server 20 executes a segmentation process (step S402) in the same manner as in step S102, thereby generating segmented lung field images. Next, the control unit 21 of the support server 20 executes a prediction process using the divided lung field images (step S403). Specifically, the prediction unit 213 of the control unit 21 inputs the created divided lung field images to a division prediction model recorded in the learning result storage unit 24, and predicts the division judgment result. Here, the negative probability and positive probability are calculated for each divided lung field in the division judgment result.

[0046] Next, the control unit 21 of the support server 20 executes a prediction process using the entire lung field image (step S404). Specifically, the prediction unit 213 of the control unit 21 inputs the feature amount for each lung field region generated in the prediction process using the divided lung field image (step S403) into the integrated prediction model recorded in the learning result storage unit 24 to calculate an integrated prediction result. In this embodiment, the prediction unit 213 generates a heat map in which a map in which the feature amount map calculated in step S403 is multiplied by a weighting value is superimposed. Then, the prediction unit 213 inputs the created heat map into the integrated prediction model recorded in the learning result storage unit 24 to predict the integrated prediction result. As the integrated judgment result, a negative probability and a positive probability are calculated.

[0047] Next, the control unit 21 of the support server 20 executes a process of outputting the judgment result (step S405). Specifically, the prediction unit 213 of the control unit 21 outputs a judgment result screen to the user terminal 10. This judgment result screen includes the judgment results (split judgment result, integrated judgment result) predicted in steps S403 and S404, and a heat map.

[0048] According to this embodiment, the following effects can be obtained. (1) In this embodiment, the control unit 21 of the support server 20 executes an image acquisition process (step S101). This makes it possible to acquire the judgment result of the image interpretation doctor. Since this image contains noise such as bones and organs that may affect the abnormality judgment on the image, machine learning that takes noise into consideration can be performed.

[0049] (2) In this embodiment, the control unit 21 of the support server 20 executes a segmentation process (step S102) for each lung field image recorded in the image information storage unit 22, a generation process of first teacher information for each lung field (step S103), and a segmentation learning process (step S104) for each segmented lung field. This makes it possible to determine the presence or absence of local abnormalities in the segmented lung field regions. For example, lung cancer has different appearances and points to be noted for each region, and by performing specialized learning for each region, it is possible to realize a judgment close to that of a radiology doctor. Furthermore, by performing machine learning after segmenting the lung fields by aligning the anatomical positions, it is possible to improve prediction accuracy.

[0050] (3) In this embodiment, the control unit 21 of the support server 20 uses "Dense Net" to execute a process of generating a feature map image in multiple layers (step S202). Here, by alternately inserting "Dense Block" and "Transition Layer", it is possible to deal with the loss of gradient when the layers are deepened. Then, it is possible to generate a feature map image with various filters applied for each layer.

[0051] (4) In this embodiment, the control unit 21 of the support server 20 executes a process of creating a heat map (step S301) using the feature map image and weighted values, a process of registering second teacher information (for integrated learning) (step S302), and a process of creating an integrated prediction model (step S303). The judgment in the divided area can be adjusted throughout. For example, when an abnormality is detected in an organ arranged symmetrically on the left and right sides of the human body, a prediction model can be generated that adjusts the judgment according to the left and right arrangement of the abnormal area. Specifically, when a suspected abnormality is detected on the upper left lung side, it can be compared with the upper right lung side to confirm the difference between the left and right sides, and the negative probability and positive probability can be calculated based on the presence or absence of a similar area on the right side. Here, if a similar area is detected on the right side, there is a possibility that it is not an abnormality.

[0052] (5) In this embodiment, the control unit 21 of the support server 20 executes image acquisition processing (step S401), prediction processing using segmented lung field images (step S403), and prediction processing using the entire lung field image (step S404). This improves the quality of diagnosis, reduces the burden on image interpretation doctors, and shortens the doctor's working hours. As a result, the burden on doctors can be shifted from primary image interpretation to tasks such as secondary image interpretation, improving the quality of comparative image interpretation. In addition, systematizing the image interpretation of lung cancer screening can contribute to eliminating regional disparities caused by the uneven distribution of highly specialized image interpretation doctors.

[0053] (6) In this embodiment, the control unit 21 of the support server 20 executes a process of outputting the determination result (step S405). In this case, by outputting a heat map or the like, it is possible to visualize targets that may be characteristics for determining an abnormality or the like.

[0054] This embodiment can be modified as follows: This embodiment and the following modifications can be combined with each other to the extent that there is no technical contradiction. In the above embodiment, the present invention is applied to the detection of lung cancer in X-ray images for lung cancer screening. The disease to which the present invention is applied is not limited to lung cancer. The present invention can also be applied to organs other than the lungs (for example, the brain and kidneys, which have a nearly symmetrical shape). In the above embodiment, the probability of being positive or negative for lung cancer is predicted. The prediction target is not limited to two types, positive and negative. For example, in the presence or absence of abnormal findings in the diagnostic information (a), the judgment results such as "no abnormal findings", "suspected lung cancer", and "abnormal findings are found and a condition requiring treatment for a disease other than lung cancer is considered" may be predicted individually. In this case, first and second teacher information including each judgment result is generated, and the probability of "no abnormal findings" is predicted by the split prediction model and the integrated prediction model, and the probability of "suspected lung cancer" is predicted.

[0055] In the above embodiment, "Dense Net" is used as the machine learning method. The machine learning method is not limited to "Dense Net", and for example, "Residual Network (ResNet)" may be used. In the above embodiment, the control unit 21 of the support server 20 executes a process of creating an integrated prediction model (step S303). In this case, a heat map for each lung field region is used as the integrated input feature. The integrated input feature is not limited to the heat map itself.

[0056] For example, as shown in FIG. 12, a heat map 600 for each lung field region and the positive probability and negative probability calculated by a division prediction model may be used as the integrated input feature. 13, an image group 530 in which each feature amount map image 520 for each lung field region is multiplied by a weighting value (W11 to W14) may be used as an integrated input feature amount. In this case, it is not necessary to use all feature amount map images, and feature amount map images with weighting values ​​equal to or greater than a reference value may be specified and used.

[0057] 14, an image group 530 in which each feature map image 520 for each lung field region is multiplied by a weight value (W11 to W14), and the positive probability and negative probability may be used as integrated input features. In this case, it is not necessary to use all feature map images, and feature map images with weight values ​​equal to or greater than a reference value may be specified and used.

[0058] In the above embodiment, the acquisition unit 211 divides an X-ray image (whole lung field image) into four lung field images from the upper left to the lower right by pattern recognition. In this pattern recognition, anatomical feature points of the lung field area are detected from the X-ray image, and the lung field area is divided based on the feature points. The division method is not limited to the use of feature points. For example, each lung field area may be divided by recognizing the shape of an organ by machine learning.

[0059] In the above embodiment, the control unit 21 of the support server 20 executes a division process (step S102). Here, the acquisition unit 211 of the control unit 21 divides the entire lung field image into an upper left lung field image, a lower left lung field image, an upper right lung field image, and a lower right lung field image by pattern recognition. Here, the number of divisions is not limited to four. In addition, the shape of the lung field may be deformed so as to align with the average shape, and the lung field region may be divided from the aligned image. [Explanation of symbols]

[0060] 10...user terminal, 20...support server, 21...control unit, 211...acquisition unit, 212...learning unit, 213...prediction unit, 22...image information storage unit, 23...teacher information storage unit, 24...learning result storage unit, 25...evaluation target storage unit.

Claims

1. a learning result storage unit that records the divided prediction model and the integrated prediction model; A judgment prediction system comprising: a control unit for evaluating an image; and a judgment prediction system for predicting a judgment result, The control unit: A plurality of divided images are generated by dividing the original image; generating first teaching information in which a determination result in the original image is associated with each of the plurality of divided images and recorded; generating a segmentation prediction model for each segmented region that predicts the determination result from the segmented image using the first teacher information, and recording the generated segmentation prediction model in the learning result storage unit; generating second teaching information by summarizing feature information for each divided region generated by inputting each divided image of the first teaching information into the division prediction model based on a position of the divided image; generating an integrated prediction model for predicting the determination result using the second teacher information, and recording the generated integrated prediction model in the learning result storage unit; A judgment prediction system characterized in that, when an evaluation target image is acquired, the judgment result is predicted using the divided prediction model and the integrated prediction model recorded in the learning result storage unit.

2. The judgment prediction system according to claim 1 , characterized in that the control unit uses feature information generated in an intermediate layer from an input layer to an output layer of the division prediction model in generating the second teaching information.

3. The system for predicting a judgment result according to claim 2 , wherein the control unit acquires a plurality of feature amount map images generated in the intermediate layer.

4. The control unit: Calculating weighting information linking the feature map image to the determination result; The judgment prediction system according to claim 3 , wherein the second teacher information is generated using the weighted information.

5. The system according to claim 4 , wherein the control unit generates a heat map using the feature amount map image and the weighting information.

6. The control unit: An organ image obtained by photographing an organ is used as the original image and the evaluation target image, 6. The diagnosis prediction system according to claim 1, wherein a diagnosis result relating to the organ image is used as the diagnosis result.

7. a learning result storage unit that records the divided prediction model and the integrated prediction model; A method for predicting a judgment result using a judgment prediction system including a control unit for evaluating an image, the method comprising: The control unit: A plurality of divided images are generated by dividing the original image; generating first teaching information in which a determination result in the original image is associated with each of the plurality of divided images and recorded; generating a segmentation prediction model for each segmented region that predicts the determination result from the segmented image using the first teacher information, and recording the generated segmentation prediction model in the learning result storage unit; generating second teaching information by summarizing feature information for each divided region generated by inputting each divided image of the first teaching information into the division prediction model based on a position of the divided image; generating an integrated prediction model for predicting the determination result using the second teacher information, and recording the generated integrated prediction model in the learning result storage unit; A judgment prediction method, characterized in that, when an evaluation target image is acquired, the judgment result is predicted using the divided prediction model and the integrated prediction model recorded in the learning result storage unit.

8. A judgment prediction program for causing a computer to function so as to output a judgment result based on an image, comprising: The method comprises: a division prediction model; and an integrated prediction model that is combined so that an output from the division prediction model is input; A plurality of divided images are generated by dividing the original image; generating first teaching information in which a determination result in the original image is associated with each of the plurality of divided images and recorded; A segmentation prediction model is generated for each segmented region by using the first teacher information to predict the determination result from the segmented image; generating second teaching information by summarizing feature information for each divided region generated by inputting each divided image of the first teaching information into the division prediction model based on a position of the divided image; an integrated prediction model generated by using the second teacher information to predict the determination result; a judgment prediction program for functioning as a means for predicting the judgment result using the division prediction model and the integrated prediction model when an evaluation target image is acquired;

Images