Program, medical information processing apparatus, and medical information processing method

Abstract

To provide a program, a medical information processor, and a medical information processing method capable of contributing to patient prognosis prediction effectively utilizing a plurality of kinds of patient information in a form of adopting medical knowledge even when the number of learning data for machine learning is not sufficient.SOLUTION: A program causes a medical information processing apparatus 11 for processing medical information to execute a procedure of executing a first prediction model for predicting a first future state from a first feature amount extracted from an affected part medical image of a patient, a procedure of executing a second prediction model for predicting a second future state from a second feature amount extracted from patient data of a patient different from the affected part medical image, and a procedure of executing a third prediction model for predicting a future patient state using the first future state, the second future state, and a predetermined parameter concerning the patient as inputs.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a program, a medical information processing device, and a medical information processing method. [Background technology]

[0002] As an example of a prediction device, prediction system, control method, and control program for predicting joint symptoms, etc., Patent Document 1 describes a prediction system including multiple prediction devices and one or more terminal devices communicably connected, in which the prediction device for a medical facility is equipped with a first prediction unit having a first prediction model that outputs first prediction information that predicts a disease of a subject at a second time point a predetermined period of time from a medical image showing the subject at a first time point, and a second prediction unit having a second prediction model that outputs third prediction information that indicates a method of intervention for the subject and the effect of the intervention from the first prediction information. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2023-169392 Summary of the Invention [Problem to be solved by the invention]

[0004] Machine learning techniques that use patient information to predict prognosis have been developed in recent years, primarily for application in medical information analysis systems that support doctors in their medical treatment.

[0005] The patient information used includes medical images such as CT and MRI, medical data such as electronic medical records, and blood test results.

[0006] When using this information to predict a patient's prognosis, it is considered more desirable in terms of accuracy and compatibility with doctors to use a system that uses two or more types of data (a multimodal system) rather than using a single type of data, such as only images or only test results. Compatibility with doctors is important because doctors often refer to multiple types of information rather than a single type of information when providing medical care.

[0007] To realize this multimodal system, a technique has been proposed in which joint symptoms and the like are predicted using both image information and basic information (examination data, etc.), as in Patent Document 1 mentioned above.

[0008] However, in the conventional techniques including the above-mentioned Patent Document 1, if the number of training data is insufficient, it is difficult to predict the prognosis of a patient with high accuracy, and therefore there is room for improvement in accuracy.

[0009] The object of the present invention is to provide a program, a medical information processing device, and a medical information processing method that can contribute to predicting patient prognosis by effectively utilizing multiple types of patient information while incorporating medical knowledge, even when the number of training data for machine learning is insufficient. [Means for solving the problem]

[0010] The present invention includes multiple means for solving the above-mentioned problems. One example is a program to be executed by a processing device that processes medical information, which causes the processing device to execute the following steps: a first prediction model that predicts a first future state from a first feature extracted from a medical image of the patient's affected area; a second prediction model that predicts a second future state from a second feature extracted from patient data of the patient that is different from the medical image of the affected area; and a third prediction model that predicts the future patient state using the first future state, the second future state, and predetermined parameters related to the patient as input. [Effects of the Invention]

[0011] According to the present invention, even when the number of training data for machine learning is insufficient, it is possible to contribute to predicting patient prognosis by effectively utilizing multiple types of patient information in a manner that incorporates medical knowledge. Problems, configurations, and effects other than those described above will become clear from the description of the following examples. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a diagram illustrating an example of a system configuration including a medical information processing apparatus according to an embodiment. [Figure 2] 1 is a flow chart illustrating an example of an outline of a medical image processing method according to an embodiment. [Figure 3] 10A and 10B are diagrams illustrating the circularity calculated from a medical image in the medical image processing apparatus according to the embodiment. [Figure 4] FIG. 4 is a partially enlarged view of FIG. [Figure 5] FIG. 10 is a diagram illustrating another example of a system configuration including a medical information processing apparatus according to an embodiment. [Figure 6] 10 is a flowchart illustrating another example of the outline of the medical image processing method according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0013] Below, we will explain a program to be executed by a processing device that processes medical information that can contribute to predicting patient prognosis by effectively utilizing multiple pairs of patient information, incorporating medical knowledge, when machine learning is applied to a medical information processing system that supports doctors' decisions and the number of learning data for machine learning is insufficient, a medical information processing device, and a medical information processing method.

[0014] In the drawings used in this specification, identical or corresponding components are denoted by the same or similar reference numerals, and repeated description of these components may be omitted.

[0015] First, an overview of the overall configuration of a system including a medical information processing device capable of suitably executing a program will be described with reference to Fig. 1. Fig. 1 is a diagram showing an example of a system configuration including a medical information processing device according to an embodiment.

[0016] As shown in FIG. 1, the system includes a medical information processing device 11, an input device 10, a medical information storage unit 12, and an output device 13.

[0017] The input device 10 is a device that receives input from an operator or a higher-level system and transmits signals related to the input to the medical information processing device 11, and is composed of a keyboard, mouse, touch panel display, etc. The medical information storage unit 12 stores medical images and other medical information and is a device that transmits and receives this information to and from the medical information processing device 11, and is composed of recording media such as HDD, SSD, and memory, and a data server of a hospital, examination center, etc. The output device 13 is a part that outputs medical information obtained from the medical information processing device 11, and is composed of a liquid crystal display, organic EL display, printer, etc.

[0018] The medical information processing device 11 is a system that predicts future patient conditions based on medical information, and includes a first medical information acquisition unit 21, a first prediction unit 22, a second medical information acquisition unit 23, a second prediction unit 24, a feature integration unit 25, and an integrated prediction unit 26.

[0019] The first medical information acquisition unit 21 is a part that acquires first medical information such as medical images of a target patient from the medical information storage unit 12 .

[0020] The first prediction unit 22 is a part that calculates first feature amounts from first medical information such as medical images of the patient's affected area acquired by the first medical information acquisition unit 21, predicts the prognosis (first future state) of the patient, and outputs the result to the feature amount integration unit 25. The first prediction unit 22 preferably executes procedures and steps for executing a first prediction model that predicts the first future state from the first feature amounts extracted from medical images of the patient's affected area.

[0021] The second medical information acquiring unit 23, like the first medical information acquiring unit 21, is a part that acquires patient data (second medical information) such as clinical information obtained from the electronic medical record and blood test results from the medical information storage unit 12.

[0022] The second prediction unit 24 is a part that, for example, calculates second feature amounts from patient data (second medical information) of a patient different from the affected area medical image acquired by the second medical information acquisition unit 23, predicts the prognosis (second future state) of the patient, and outputs the result to the feature amount integration unit 25. The second prediction unit 24 preferably executes the procedure / steps of executing a second prediction model that predicts the second future state from the second feature amounts extracted from the patient data of a patient different from the affected area medical image.

[0023] The feature integration unit 25 acquires the prognosis prediction results output from each of the first prediction unit 22 and the second prediction unit 24, and calculates predetermined parameters related to the patient from information about the patient acquired from the medical information storage unit 12, and integrates the two prognosis prediction results in a predetermined manner.

[0024] The integrated prediction unit 26 is a part that predicts the patient's prognosis (future patient condition) using the integrated prognosis prediction result, and outputs the predicted patient's prognosis to the output device 13. Furthermore, this integrated prediction unit 26 can output a display signal to display on the output device 13, in addition to the patient's prognosis (future patient condition), at least one of the first future state predicted by the first prediction unit 22 and the second future state predicted by the second prediction unit 24.

[0025] The future patient condition, which is the output result obtained by the medical information processing device 11 of the present invention, is, for example, graphs or numerical values ​​such as recurrence rate, survival rate, mortality rate, or incidence rate of adverse events, or in the case of cancer, information on progression such as tumor size, and is displayed in a state linked to the patient ID. At this time, it may be useful for the doctor to understand the future patient condition to also display the first future condition predicted by the first prediction unit 22 and the second future condition predicted by the second prediction unit 24, so they can be displayed together.

[0026] The feature integration unit 25 and the integrated prediction unit 26 preferably act as the entities that execute the procedures and steps of executing a third prediction model that predicts the future patient state using the first future state, the second future state, and predetermined parameters related to the patient as input, and the integrated prediction unit 26 preferably acts as the entities that execute the procedures and steps of causing the medical information processing device 11 to execute a procedure of outputting to the output device 13 a display signal that displays at least one of the first future state and the second future state.

[0027] In these feature integration unit 25 and integrated prediction unit 26, when predicting a patient's prognosis, the future patient state can be predicted by integrating the first future state and the second future state based on predetermined parameters related to the patient.

[0028] The components, functions, processing units, processing means, etc. constituting the above-described medical information processing device 11 can be partly or entirely realized in hardware by, for example, designing them as integrated circuits. Also, the components, functions, etc. can be realized in software by a processor interpreting and executing a program that realizes each function.

[0029] Information such as programs, tables, files, etc. that realize each function in the above-mentioned medical information processing device 11 can be stored in a memory, a recording device such as a hard disk or SSD (Solid State Drive), or a recording medium such as an IC card, SD card, or DVD.

[0030] In addition, the control lines and information lines shown are those that are considered necessary for the explanation, and do not necessarily show all the control lines and information lines in the product. In reality, it can be considered that almost all components are interconnected.

[0031] Next, an example of the flow of a medical information processing method suitably executed by the medical information processing device 11 having the system configuration shown in Fig. 1 will be described with reference to Fig. 2. Fig. 2 is a flow for explaining an example of an outline of a medical image processing method according to an embodiment.

[0032] Here, we will explain the case where medical images such as CT images and X-ray images are input as the first medical information, and numerical values ​​such as blood test results are input as the second medical information.

[0033] 2, first, the first medical information acquisition unit 21 acquires a medical image from the medical information storage unit 12 in response to an instruction from the system or an operator (step S101). Next, the first prediction unit 22 calculates a medical image feature (first feature) from the acquired medical image and predicts the prognosis (first future state) of the patient (step S102).

[0034] In addition, the second medical information acquisition unit 23 acquires medical information from the medical information storage unit 12 in response to instructions from the system or the operator (step S103), and the second prediction unit 24 calculates a second feature from the acquired medical information and predicts the patient's prognosis (second future condition) (step S104).

[0035] Here, steps S103 and S104 may be executed before steps S101 and S102, or may be executed simultaneously if the system is capable of parallel processing.

[0036] Next, the feature amount integrating unit 25 acquires patient information from the medical information storage unit 12 and calculates predetermined parameters related to the patient (step S105).

[0037] Furthermore, the feature integration unit 25 integrates the first future state and the second future state obtained from the first prediction unit and the second prediction unit, respectively, using a preset method based on predetermined parameters related to the patient to calculate an integrated feature (step S106).

[0038] Thereafter, the integrated prediction unit 26 predicts the prognosis of the patient from the integrated feature calculated by the feature integration unit 25 (step S107). Thereafter, the predicted result is output to the output device 13 (step S108). The feature here refers to a value representing the feature of the data to be analyzed, i.e., the first and second medical information.

[0039] Next, an example of a method for obtaining image feature amounts used when the first prediction unit 22 predicts the first future state, which is the prognosis of the patient, will be described.

[0040] The medical information used in the first prediction unit 22 is mainly medical images such as CT images and X-ray images.

[0041] The feature quantity representing the characteristics of the medical information can be a value calculated based on a predefined formula, such as features related to the shape of the lesion obtained from the image, such as circularity or edge sharpness, or information related to the tendency of the brightness distribution, such as contrast or brightness variation. In this way, the first feature quantity can include a numerical value related to the tendency of the brightness distribution of the lesion.

[0042] An example of calculation of the circularity is shown in Figures 3 and 4. Figure 3 is a diagram for explaining the circularity calculated from a medical image in the medical information processing apparatus according to the embodiment, and Figure 4 is a partial enlarged view of Figure 3.

[0043] The circularity of the lesion 30 in the medical image shown in FIG. 3 can be determined by the ratio of the minor diameter 33 to the major diameter 35 as shown in FIG. 4, for example, by dividing the major diameter 35 by the minor diameter 33.

[0044] In addition to this method, other methods for identifying the area of ​​a lesion from a medical image include obtaining and using the position and size from medical information, identifying the area from image information such as brightness based on pre-set rules (rule-based), and using machine learning techniques.

[0045] The first prediction unit 22 uses these feature amounts to output a prognosis prediction result for the patient. The output numerical value may be, for example, a prediction result calculated using a rule-based method or a prediction result calculated using a machine learning method including a support vector machine (SVM) or a deep learning method.

[0046] The predicted results are assumed to be probabilities (decimals between 0 and 1) of pre-set prognosis definitions, such as the 5-year survival rate, the possibility of recurrence after treatment, the incidence of adverse events after treatment, etc. Alternatively, these probabilities can be converted into binary values ​​based on a pre-set threshold.

[0047] In addition, there is also a method of predicting prognosis from an input image using a deep learning technique represented by a CNN (Convolutional Neural Network). In this case, the prognosis prediction result output from the first prediction unit 22 can be the prognosis prediction result of the CNN, or the output value of the layer just before the final layer of the CNN can be used as the prognosis prediction result.

[0048] Next, an example of a method for obtaining clinical information feature quantities used when the second prediction unit 24 predicts the second future state, which is the prognosis of the patient, will be described.

[0049] The medical information input to the second prediction unit 24 is a number of numerical values ​​calculated according to pre-set criteria, such as numerical values ​​about the patient extracted from blood test results and electronic medical records (binary values ​​such as height, weight, age, and whether or not there is a medical history, Pack Years = (number of cigarettes smoked per day / 20 cigarettes) x number of years of smoking), and Performance Status (an evaluation index of the patient's overall condition, indicating how well the patient is able to carry out daily activities).

[0050] The feature for these multiple numerical values ​​(numerical value group) may be, for example, the numerical value group itself, a normalized numerical value group obtained by normalizing the numerical value group based on a preset criterion, or a numerical value group extracted from the numerical value group based on a preset rule. In this way, the second feature may include a numerical value representing the patient's condition calculated based on a preset criterion.

[0051] The second prediction unit 24 outputs a patient prognosis prediction result using any of the above-mentioned groups of numerical values. As with the first prediction unit 22, the prediction method may be, for example, a rule-based method or a machine learning method such as SVM, and the output numerical values ​​are the calculated prediction results.

[0052] The predicted results are assumed to be probabilities (decimals between 0 and 1) of pre-set prognosis definitions, such as the 5-year survival rate, the possibility of recurrence after treatment, the incidence of adverse events after treatment, etc. Alternatively, these probabilities can be converted into binary values ​​based on a pre-set threshold.

[0053] In this case, a method using deep learning techniques such as CNN can also be used. In this case, the prognosis prediction result output from the second prediction unit 24 can be the prognosis prediction result of CNN, or the output value of the layer just before the final layer of CNN can be used as the prognosis prediction result.

[0054] Next, an example of a method for calculating the integrated feature amount in the feature amount integration unit 25 will be described.

[0055] The following formula (1) is an example of a feature integration formula selected in step S106. Here, it is assumed that the prognosis of a lung cancer patient is predicted.

[0056]

number

[0057] In the above formula (1), Vol tumor is the tumor volume, Vol glass is the volume of the ground-glass opacity inside the tumor, α is a hyperparameter, f(p) is the integrated feature of patient p, Feature img (p) is the image feature of patient p, Feature info (p) indicates the clinical information feature of patient p, and Vol glass / Vol tumor and Cp are predetermined parameters related to the patient, and the existence of Cp and (1-Cp) as terms in equation (1) provides weighting, and the hyperparameter α is a value specific to the model.

[0058] In this manner, the predetermined patient parameters may include values ​​obtained from image processing of the patient's medical images, or may include values ​​obtained based on information obtained from the patient's medical records, and the third prediction model may integrate the first future state and the second future state by weighting based on the predetermined patient parameters.

[0059] Here, the image feature in equation (1) img (p) and clinical information feature info (p) is expressed as a one-dimensional or higher-dimensional array of positive numbers or decimals. Therefore, the calculated integrated feature is also an array of one-dimensional or higher dimensions.

[0060] In addition, "&" in equation (1) represents concatenation. For example, img (p) is e img A one-dimensional array with elements, clinical information feature info (p) is e info When it is a one-dimensional array with elements, the integrated feature is (e img +e info ) elements.

[0061] Furthermore, the "tumor" referred to here is a shadow that appears in a medical image, which is the primary medical information. In CT images of the chest area capturing the lungs, the tumor often appears brighter than the surrounding air area, and in other organs, it often appears as an area with a different brightness value from tissues other than the tumor.

[0062] In the case of a lung cancer patient, the patient information acquired in step S105 is the volume of the ground-glass opacity and the volume of the tumor. Ground-glass opacity refers to a semi-transparent opacity (enough to allow the tissue behind it to be seen) on a CT image of the lung field. The volume may be calculated from brightness information, etc., using a preset method from the medical image, or if the patient information is recorded in, for example, the medical record or image interpretation information, that value may be used. α is a value that has been appropriately adjusted based on prior consideration.

[0063] According to equation (1), when the ratio of the volume of ground-glass opacity inside the tumor is large, the image feature img In the opposite case, the clinical information feature Feature info This is an integration method that emphasizes (p). This is based on the knowledge that, for example, when the proportion of ground-glass opacities in a tumor is high, the cancer is less advanced, and vice versa, the cancer is often more advanced. In preliminary studies, when the accuracy of prognosis prediction based on clinical information alone was compared with that based on images alone, the integration method is assumed to be effective when the accuracy of clinical information is higher when the cancer stage is high, and when the accuracy of images is higher when the cancer stage is low.

[0064] Another feature integration method will be described. Let us consider a case where a breast cancer patient is assumed and mammography images are used as medical images to predict the prognosis.

[0065] The breast is composed of mammary glandular tissue and fatty tissue, but when photographed with a mammogram, the mammary gland appears as high brightness (white), the fat as low brightness (black), and the tumor as high brightness (white). Generally, the tumor area is identified from the difference in brightness between the fat, which is the main background, and the tumor, and image features are calculated from this.

[0066] Some patients have the characteristic of having a large number of mammary glands (high-density mammary glands). In patients with high-density mammary glands, the mammogram will likely appear bright overall, with the tumor area hidden by the shadow of fat, making it difficult to identify the characteristics of the lesion area. For this reason, in the case of patients with high-density mammary glands, an integration method is used that places more emphasis on clinical information features than in patients without high-density mammary glands.

[0067] For patients with dense breast tissue, the patient background information (electronic medical records or radiology reports) may contain information such as "dense breast tissue" or "dense," which can be used as patient information. Alternatively, the degree of breast tissue density can be estimated separately from the first prediction section and used as patient information.

[0068] In step S106 executed by the feature amount integration unit 25, not only a formula set in advance but also machine learning can be used. For example, α used in the above formula (1) can be obtained by machine learning.

[0069] By using machine learning for this process, integration can be performed in a more flexible manner, which is expected to stabilize the performance of prognosis prediction.

[0070] Alternatively, instead of dividing the process into two units, the feature integration unit 25 and the integrated prediction unit 26, and separately performing feature integration and prognosis prediction as in steps S106 and S107, it is also possible to replace the process with a configuration in which image features, clinical information features, and patient information are input and a prognosis prediction is output. An example of the system configuration in this case is shown in Fig. 5, and an example of the information processing flow is shown in Fig. 6. Fig. 5 is a diagram showing another example of the system configuration including the medical image processing device according to the embodiment, and Fig. 6 is a flow chart for explaining another example of the outline of the medical image processing method according to the embodiment.

[0071] As shown in FIG. 5, another embodiment of the present system includes a medical information processing device 11A, an input device 10, a medical information storage unit 12, and an output device 13.

[0072] The medical information processing device 11A is a system that predicts future patient conditions based on medical information, and has a first medical information acquisition unit 21, a first prediction unit 22, a second medical information acquisition unit 23, a second prediction unit 24, and a prediction unit 27.

[0073] In FIG. 5, the input device 10, medical information storage unit 12, output device 13, first medical information acquisition unit 21, first prediction unit 22, second medical information acquisition unit 23, and second prediction unit 24 are the same as those in FIG.

[0074] The prediction unit 27 acquires the prognosis prediction results output from the first prediction unit 22 and the second prediction unit 24, and determines predetermined parameters related to the patient from information about the patient acquired from the medical information storage unit 12.The prediction unit 27 predicts the patient's prognosis (future patient condition) using these prognosis prediction results and the predetermined parameters related to the patient, and outputs the predicted patient's prognosis to the output device 13.

[0075] In addition, the prediction unit 27 can output a display signal to display on the output device 13, in addition to the patient's prognosis (future patient condition), either the first future condition predicted by the first prediction unit 22 or the second future condition predicted by the second prediction unit 24.

[0076] In this case, the prediction unit 27 preferably executes the procedure / steps to execute a third prediction model that predicts the future patient state using the first future state, the second future state, and predetermined parameters related to the patient as input.

[0077] Regarding the processing flow, steps S101 to S105 and step S108 shown in Fig. 6 are the same as the steps in Fig. 2. In Fig. 6, steps S106 and S107 in Fig. 2 can be replaced with step S201 that executes prognosis prediction using image prediction features, clinical information prediction features, and patient information.

[0078] Next, the effects of this embodiment will be described.

[0079] The program of the present embodiment described above causes a medical information processing device 11 that processes medical information to execute the following steps: a first prediction model that predicts a first future state from a first feature extracted from a medical image of the patient's affected area; a second prediction model that predicts a second future state from a second feature extracted from patient data of a patient other than the medical image of the affected area; and a third prediction model that predicts the future patient state using the first future state, the second future state, and specified parameters related to the patient as input.

[0080] By incorporating knowledge into the prediction model for each piece of information in this way, even when a large amount of data covering the state of a disease is not available as learning data or is not accumulated enough, it is possible to provide a medical information processing device and a medical information processing method that can contribute to predicting patient prognosis by effectively utilizing multiple types of patient information in a way that incorporates medical knowledge, whereas there is no guarantee that the designed prediction system will be consistent with medical knowledge, as in conventional technology.

[0081] In addition, the third prediction model predicts the future patient state by integrating the first future state and the second future state based on specified parameters related to the patient, thereby enabling more accurate prediction of the patient's prognosis.

[0082] Furthermore, by making the first feature amount include a numerical value relating to the tendency of the brightness distribution of the lesion, by making the second feature amount include a numerical value representing the patient's condition calculated according to preset criteria, and by making the predetermined parameters related to the patient include numerical values ​​obtained from image processing of the patient's medical images or numerical values ​​obtained based on information obtained from the patient's medical information records, it is possible to realize prognosis prediction that is more tailored to the patient's condition.

[0083] In addition, by having the medical information processing device 11 execute a procedure to output a display signal to the output device 13 to display at least one of the first future state and the second future state, the basis for calculating the requested future prediction for the patient can also be understood, and its reliability can also be determined, contributing to providing more appropriate medical care to patients.

[0084] <Other> The present invention is not limited to the above-described embodiment, and various modifications and applications are possible. The above-described embodiment has been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to having all of the described configurations.

[0085] The embodiment of the present invention may be in the following form.

[0086] (1) A program to be executed by a processing device that processes medical information, the program causing the processing device to execute the following steps: a first prediction model that predicts a first future state from a first feature extracted from a medical image of a patient's affected area; a second prediction model that predicts a second future state from a second feature extracted from patient data of the patient that is different from the medical image of the affected area; and a third prediction model that predicts the future patient state using the first future state, the second future state, and specified parameters related to the patient as input.

[0087] (2) In the program described in (1), the third prediction model predicts the future patient state by integrating the first future state and the second future state based on predetermined parameters related to the patient.

[0088] (3) In the program according to (1) or (2), the first feature amount includes a numerical value relating to the tendency of the brightness distribution of the lesion.

[0089] (4) In the program according to any one of (1) to (3), the second feature amount includes a numerical value representing the condition of the patient calculated according to a preset criterion.

[0090] (5) In the program according to any one of (1) to (4), the predetermined parameters relating to the patient include numerical values ​​obtained from image processing of medical images of the patient.

[0091] (6) In the program described in any one of (1) to (5), the predetermined parameters related to the patient include numerical values ​​obtained based on information obtained from the patient's medical information record.

[0092] (7) In the program described in any one of (2) to (6), the third prediction model integrates the first future state and the second future state by weighting based on a predetermined parameter related to the patient.

[0093] (8) In the program described in any one of (1) to (7), the processing device is caused to execute a procedure of outputting a display signal to a display device to display at least one of the first future state and the second future state. [Explanation of symbols]

[0094] 10...Input device 11, 11A...Medical information processing device (processing device) 12...Medical information storage unit 13...Output device (display device) 21…First Medical Information Acquisition Department 22...First prediction unit (first prediction model calculation unit) 23…Second Medical Information Acquisition Department 24...Second prediction unit (second prediction model calculation unit) 25...Feature integration unit (third prediction model calculation unit) 26... Integrated prediction unit (third prediction model calculation unit) 27...Prediction unit (third prediction model calculation unit) 30...Affected area 33…Short diameter 35...long diameter

Claims

1. A program to be executed by a processing device that processes medical information, a step of executing a first prediction model that predicts a first future state from a first feature amount extracted from a medical image of the affected part of a patient; a step of executing a second prediction model that predicts a second future state from a second feature amount extracted from patient data of the patient different from the affected area medical image; and executing a third prediction model that predicts a future patient state using the first future state, the second future state, and predetermined parameters related to the patient as inputs. program.

2. 2. The program according to claim 1, The third prediction model predicts the future patient state by integrating the first future state and the second future state based on predetermined parameters related to the patient. program.

3. 2. The program according to claim 1, The first feature amount includes a numerical value related to the tendency of the brightness distribution of the lesion. program.

4. 2. The program according to claim 1, The second feature amount includes a numerical value representing the condition of the patient calculated based on a preset standard. program.

5. 2. The program according to claim 1, The predetermined parameters related to the patient include values ​​obtained from image processing of the medical images of the patient. program.

6. 2. The program according to claim 1, The predetermined parameters related to the patient include values ​​obtained based on information obtained from the patient's medical information record. program.

7. 3. The program according to claim 2, In the third prediction model, the first future state and the second future state are integrated by weighting based on predetermined parameters related to the patient. program.

8. 2. The program according to claim 1, and causing the processing device to execute a procedure of outputting a display signal to a display device to display at least one of the first future state and the second future state. program.

9. A medical information processing device that predicts a future patient condition based on medical information, a first prediction model calculation unit that predicts a first future state from a first feature amount extracted from a medical image of the affected part of a patient; a second prediction model calculation unit that predicts a second future state from a second feature amount extracted from patient data of the patient different from the affected area medical image; a third prediction model calculation unit that predicts a future patient state using the first future state, the second future state, and predetermined parameters related to the patient as inputs; Medical information processing equipment.

10. A method for processing medical information, comprising: Executing a first prediction model to predict a first future state from a first feature extracted from a medical image of the affected area of ​​a patient; Executing a second prediction model to predict a second future state from second feature amounts extracted from patient data of the patient different from the affected area medical image; and executing a third prediction model to predict a future patient state using the first future state, the second future state, and predetermined parameters related to the patient as inputs. How medical information is processed.

Images