Determination system, determination method, and program

Abstract

This determination system is provided with: a storage device that stores a trained model obtained on the basis of a whole body parameter indicating the state of the whole body, an ophthalmologic parameter indicating the state of the eye, and a correct answer label indicating the possibility of suffering from sleep apnea syndrome; and a determination unit that uses the whole body parameter of a subject, the ophthalmologic parameter of the subject, and the trained model to determine the possibility that the subject is suffering from sleep apnea syndrome. The trained model may include at least one of: an extra tree classifier; a logistic regression function unit; a k-neighborhood function unit; a naïve Bayes classifier; a decision tree; a support vector machine; a radial basis function support vector machine; a Gaussian process classifier; a multilayer perceptron function unit; a ridge classifier; a random forest function unit; a secondary discrimination analyzer; an adaptive boosting function unit; a gradient boosting classifier; a linear discrimination analyzer; an extreme gradient boosting; and a gradient boosting function unit.

Description

Determination system, determination method, and program

[0001] This application claims priority to international application PCT / JP2023 / 029855 filed on August 18, 2023, the contents of which are incorporated herein by reference.

[0002] Sleep apnea syndrome (SAS) is a disease that can cause vascular endothelial damage, arteriosclerosis, hypertension, heart failure, etc., and has also been linked to glaucoma. Therefore, there is a need for a simple method to determine whether a subject may be suffering from sleep apnea syndrome. By making a simple determination based on ophthalmological parameters, it becomes possible to efficiently connect patients to specialists in sleep apnea syndrome.

[0003] Patent No. 5979359

[0004] However, determining whether a subject has sleep apnea syndrome requires a sophisticated device, such as a device that analyzes the frequency of the subject's snoring, making it difficult to simply determine whether a subject has sleep apnea syndrome.

[0005] In view of the above circumstances, the present invention aims to provide a determination system, a determination method, and a program that can more easily determine the possibility of suffering from sleep apnea syndrome.

[0006] One aspect of the present invention is a determination system comprising: a storage device that stores a trained model obtained based on whole-body parameters that represent the condition of the entire body, ophthalmological parameters that represent the condition of the eyes, and a correct label that represents the possibility of the subject suffering from sleep apnea syndrome; and a determination unit that determines the possibility of the subject suffering from sleep apnea syndrome using the whole-body parameters of the subject, the ophthalmological parameters of the subject, and the trained model.

[0007] One aspect of the present invention is a determination method executed by a determination system having a storage device that stores a trained model, wherein the trained model is obtained based on whole-body parameters that represent the condition of the entire body, ophthalmic parameters that represent the condition of the eyes, and a correct label that represents the possibility of the subject suffering from sleep apnea syndrome, and the determination method includes a step of determining the possibility that the subject suffers from sleep apnea syndrome using the whole-body parameters of the subject, the ophthalmic parameters of the subject, and the trained model.

[0008] One aspect of the present invention is a program for causing a computer to function as a judgment system comprising: a storage device that stores a trained model obtained based on whole-body parameters that represent the condition of the entire body, ophthalmic parameters that represent the condition of the eyes, and a correct label that represents the possibility of the subject suffering from sleep apnea syndrome; and a judgment unit that uses the whole-body parameters of the subject, the ophthalmic parameters of the subject, and the trained model to judge the possibility that the subject suffers from sleep apnea syndrome.

[0009] According to the present invention, it is possible to more easily determine the possibility of suffering from sleep apnea syndrome.

[0010] 1 is a diagram illustrating an example of the configuration of a determination system in an embodiment; FIG. 2 is a diagram illustrating examples of whole-body parameters in an embodiment; FIG. 3 is a diagram illustrating an example of the prevalence of sleep apnea syndrome for patients with open-angle glaucoma in an embodiment; FIG. 4 is a diagram illustrating a comparative example between the snoring awareness rate and the apnea indication rate in an embodiment; FIG. 5 is a diagram illustrating an example of a known visual field sector map in an embodiment; FIG. 6 is a diagram illustrating examples of ophthalmological parameters in an embodiment; FIG. 7 is a diagram illustrating an example of a receiver operating characteristic curve of an extra-tree classifier in an embodiment; FIG. 8 is a diagram illustrating an example of a SHAP value representing the influence of a feature variable in an embodiment; FIG. 9 is a diagram illustrating an example of an operation screen in an embodiment; FIG. 10 is a flowchart illustrating an example of the operation of a learning device in an embodiment; FIG. 11 is a flowchart illustrating an example of the operation of a determination device in an embodiment; FIG. 12 is a diagram illustrating an example of the configuration of a determination system in a modified embodiment; FIG. 13 is a diagram illustrating an example of an average receiver operating characteristic curve with variability for a trained model trained using predetermined original training data in an embodiment;

[0011] An embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 is a diagram showing an example of the configuration of a determination system 1a according to the embodiment. The determination system 1a is a system for determining the possibility that a subject suffers from sleep apnea syndrome. For example, a high proportion of glaucoma patients suffer from sleep apnea syndrome. Therefore, the determination system 1a determines the possibility that a subject suffers from sleep apnea syndrome based on parameters that represent at least the subject's eye condition.

[0012] The determination system 1a includes a learning device 10, a determination device 20, and a communication line 30. The learning device 10 and the determination device 20 may be integrated. That is, the determination device 20 may include the learning device 10.

[0013] The learning device 10 includes a communication unit 11, a storage device 12, a storage unit 13, an input unit 14, and a learning unit 15. The learning unit 15 includes a learning control unit 151 and a learning processing unit 152.

[0014] The determination device 20 includes a communication unit 21, a storage device 22, a storage unit 23, an input unit 24, a determination unit 25, and an output unit 26. The determination unit 25 includes a determination control unit 251 and a determination processing unit 252.

[0015] Some or all of the functional units of the learning device 10 and the determination device 20 are realized as software by a processor such as a central processing unit (CPU) executing a program stored in a storage unit having a non-volatile storage medium (non-transitory storage medium). The program may be recorded on a computer-readable storage medium. Examples of computer-readable storage media include portable media such as flexible disks, magneto-optical disks, read-only memories (ROMs), and compact disc read-only memories (CD-ROMs), hard disks built into computer systems, and storage devices such as solid-state drives.

[0016] Some or all of the functional units of the learning device 10 and the judgment device 20 may be realized using hardware including electronic circuits (electronic circuits or circuitry) using, for example, an LSI (Large Scale Integrated circuit), an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), or an FPGA (Field Programmable Gate Array).

[0017] First, the learning device 10 will be described. The communication unit 11 communicates with the communication unit 21 via the communication line 30. The communication unit 11 transmits a trained model stored in the storage device 12 to the communication unit 21. The trained model is realized using a neural network having at least one of, for example, an extra trees classifier, a logistic regression function unit, a k-nearest neighbor function unit, a naive Bayes classifier, a decision tree, a support vector machine, a radial basis function support vector machine, a Gaussian process classifier, a multilayer perceptron function unit, a ridge classifier, a random forest function unit, a quadratic discriminant analyzer, an adaptive boosting function unit, a gradient boosting classifier, a linear discriminant analyzer, extreme gradient boosting, and a gradient boosting function unit.

[0018] The storage device 12 associates learning data (explanatory variables) with correct labels (objective variables) and stores them as training data. The learning data includes parameters representing the condition of the whole body (hereinafter referred to as "whole body parameters") and parameters representing the condition of the eyes (hereinafter referred to as "ophthalmological parameters"). The correct labels are correct information indicating the possibility of suffering from sleep apnea syndrome. The correct labels are expressed, for example, as percentages. The storage device 12 pre-stores a program for causing the computer to function as a learning device.

[0019] A program for causing the computer to function as a learning device is loaded into the memory unit 13 from the storage device 12. The input unit 14 is a functional unit (operation unit) that accepts operational input, and is, for example, a keyboard, a mouse, or a touch panel.

[0020] In the learning stage of machine learning, the learning control unit 151 acquires teacher data. The teacher data is input from the input unit 14, for example, in response to a user's operation. The teacher data may be acquired via communication from an external communication device (not shown). The teacher data may be acquired from a recording medium (not shown) connected to the learning device 10. Predetermined preprocessing may be performed on the teacher data. The predetermined preprocessing is, for example, a process of calculating the body mass index (BMI) based on the subject's height and weight. The learning processing unit 152 performs a learning process (supervised learning) on ​​the teacher data using a machine learning method (e.g., backpropagation). The learning control unit 151 or the learning processing unit 152 records the trained model obtained by the learning process in the storage device 12.

[0021] Next, the determination device 20 will be described. The communication unit 21 communicates with the communication unit 11 via the communication line 30. The communication unit 21 acquires, for example, a trained model obtained by the learning process by the learning processing unit 152 from the communication unit 11. The communication unit 21 records the trained model in the storage device 22. The storage device 22 stores in advance a program for causing a computer to function as a determination device. The program for causing a computer to function as a determination device is loaded from the storage device 22 into the storage unit 23.

[0022] The input unit 24 is a functional unit (operation unit) that accepts operation input, and is, for example, a keyboard, a mouse, or a touch panel. The input unit 24 acquires whole-body parameters (characteristic variables of the whole body of the subject) and ophthalmologic parameters (characteristic variables of the eye of the subject) of the subject in response to operations by a user.

[0023] In the inference stage (determination stage) after the learning stage of machine learning, the determination control unit 251 acquires whole-body parameters and ophthalmic parameters of the subject. The determination control unit 251 may perform predetermined preprocessing on the whole-body parameters and ophthalmic parameters. The determination control unit 251 inputs the acquired whole-body parameters and ophthalmic parameters to the determination processing unit 252 as explanatory variables.

[0024] The determination processing unit 252 performs a determination process (inference process) of the possibility that the subject is suffering from sleep apnea syndrome using the input whole-body parameters and ophthalmologic parameters and the trained model. The determination control unit 251 or the determination processing unit 252 displays the determination result of the determination processing unit 252 on the output unit 26. The determination result (the possibility that the subject is suffering from sleep apnea syndrome) is expressed, for example, as a percentage.

[0025] The output unit 26 includes a display device such as a liquid crystal display. The output unit 26 displays, for example, an operation screen. The output unit 26 displays, for example, the determination result of the determination processing unit 252. The output unit 26 may include an audio output unit. The output unit 26 may output the determination result of the determination processing unit 252 by audio.

[0026] Next, whole-body parameters will be described as an example of explanatory variables. Fig. 2 is a diagram showing examples of whole-body parameters in an embodiment. The whole-body parameters include, for example, age, sex (male), body mass index (BMI), smoking, high blood pressure, diabetes, dyslipidemia, heart disease, migraine, thyroid disease, sensitivity to cold, history of cancer, systolic blood pressure, diastolic blood pressure, pulse rate, apnea-hypopnea index (AHI), n (n is, for example, 4)% oxygen desaturation index (ODI), oxidative stress (d-ROMs), antioxidant capacity (biological antioxidant potential) (BAP), snoring (subjective), apnea (noted by family), and a history of diagnosis of sleep apnea syndrome (SAS).

[0027] In FIG. 2, a non-"sleep apnea syndrome" group (non-SAS group), a "sleep apnea syndrome" group (SAS group), a P-value (Probability-value), and an adjusted P-value are associated with whole-body parameters. The P-value is a value compared with a significance level. If "P-value < significance level" holds, it is determined that there is a significant difference in the whole-body parameter associated with that P-value (the whole-body parameter is an SAS factor). The adjusted P-value is a P-value adjusted according to age and gender.

[0028] When comparing the non-"sleep apnea syndrome" group with the "sleep apnea syndrome" group, as shown in the comparison results between the adjusted P value and the significance level, high BMI, a history of hypertension, systolic blood pressure, diastolic blood pressure, low antioxidant capacity, and apnea (noted by family members) are SAS factors in systemic parameters.

[0029] 3 is a diagram showing an example of the prevalence of sleep apnea syndrome in patients with open-angle glaucoma (OAG) according to an embodiment. As shown in FIG. 3, approximately 30% of patients with open-angle glaucoma have untreated sleep apnea syndrome. Thus, the prevalence of sleep apnea syndrome is high among patients with open-angle glaucoma.

[0030] FIG. 4 is a diagram showing a comparative example of the snoring awareness rate and the apnea indication rate in an embodiment. When comparing the non-"sleep apnea syndrome" group with the "sleep apnea syndrome" group, the snoring awareness rate is approximately 30% in both groups. Thus, no significant difference in the snoring awareness rate is observed between the non-"sleep apnea syndrome" group and the "sleep apnea syndrome" group. In contrast, the indication rate of apnea (notified by family members) is significantly higher in the "sleep apnea syndrome" group. However, the indication rate of apnea (notified by family members) in the "sleep apnea syndrome" group is low at 24%. Therefore, sleep apnea syndrome cannot be suspected based solely on subjective symptoms of snoring and apnea.

[0031] Next, ophthalmological parameters will be described as an example of explanatory variables. Fig. 5 is a diagram showing an example of a known visual field sector map (Garway-Heath map) in an embodiment. The visual field sector map is a map of evaluation values ​​of visual field sectors that anatomically correspond to four predetermined regions (superior, temporal, inferior, and nasal) around the optic nerve head (ONH). The evaluation value of glaucomatous visual field defects is, for example, total deviation (TD).

[0032] Figure 5 shows an example of a visual field sector map for the right eye. The rightmost regions in Figure 5 are nasal regions. The central regions of the visual field sector map are temporal regions. The temporal regions are further divided into an upper region (above the center) and a lower region (below the center). The average total deviation of the visual field (hereinafter referred to as the "average total deviation") is calculated for each of the upper, upper-center, lower-center, and lower regions.

[0033] FIG. 6 is a diagram illustrating an example of ophthalmic parameters (comparison between ophthalmic parameters of a non-sleep apnea syndrome group and a sleep apnea syndrome group) in an embodiment. As an example, the number of 4% oxygen desaturation index (4% ODI) events was measured. Based on this measurement result, the subjects were divided into a sleep apnea syndrome group (15 or more events per hour) and a non-sleep apnea syndrome group (less than 15 events per hour). The ophthalmic parameters include, for example, best corrected visual acuity, intraocular pressure, central corneal thickness, axial length, peripapillary retinal nerve fiber layer thickness (CpRNFLT) (superior, temporal, inferior), mean blur ratio (MT) of the optic disc tissue (superior, temporal, inferior), number of components of glaucoma eye drops, mean deviation (MD), and mean total deviation (superior, central superior, central inferior, inferior).

[0034] The average blur rate of the optic disc tissue portion represents the blood flow velocity of the capillaries in the optic disc. Since there is a correlation between the refraction of the eye and the axial length, the refraction of the eye may be used as the ophthalmic parameter instead of the axial length. The refraction of the eye may be expressed, for example, in terms of the prescription of glasses. Since eye drops for glaucoma reduce intraocular pressure, the intraocular pressure may be used as the ophthalmic parameter instead of the number of ingredients in the eye drops. As such, many ophthalmic parameters can be obtained relatively easily.

[0035] 6, a non-"sleep apnea syndrome" group (non-SAS group), a "sleep apnea syndrome" group (SAS group), a P value, and an adjusted P value are associated with ophthalmological parameters. If "P value < significance level" is satisfied, it is determined that there is a significant difference in the ophthalmological parameter associated with the P value (the ophthalmological parameter is an SAS factor).

[0036] When comparing the non-"sleep apnea syndrome" group with the "sleep apnea syndrome" group, as shown in the comparison results between the adjusted P value and the significance level, the SAS factors in the ophthalmological parameters were a short axial length, a thin temporal peripapillary retinal nerve fiber layer thickness (CpRNFLT), a small number of ingredients in the glaucoma eye drops, and a decrease in sensitivity in the central inferior visual field ("small central inferior mean total deviation").

[0037] 7 is a diagram illustrating an example of a receiver operating characteristic (ROC) curve for an extra-trees classifier in an embodiment. The area under the ROC curve (AUC) of a trained model with an extra-trees classifier is the largest compared to the areas under the ROC curve of trained models with other classifiers or functional units. Specifically, as an example, when systemic parameters ("BMI," "male," "history of hypertension," "age," and "apnea (notified by family)") and ophthalmological parameters ("decreased sensitivity in the central inferior visual field," "number of ingredients in glaucoma eye drops," and "axial length") are input to a trained model, and the trained model includes an extra-trees classifier, the area under the ROC curve is 0.94.

[0038] 8 is a diagram showing examples of SHAP (Shapley Additive exPlanations) values ​​that represent the influence of feature variables in an embodiment. The wider the distribution of the SHAP values ​​shown on the horizontal axis, the more influential the parameter is (the SAS factor). BMI, being male, having a history of hypertension, age, and apnea (noted by family members) are examples of systemic parameters that influence the assessment.

[0039] In addition, reduced sensitivity in the infracentral visual field (small infracentral mean total deviation), the number of ingredients in glaucoma eye drops, and axial length are examples of ophthalmic parameters that affect the assessment. Here, the positive and negative SHAP values ​​are clearly determined depending on the value of the infracentral mean total deviation (feature variable). Therefore, it is possible to more accurately assess the possibility that a subject suffers from sleep apnea syndrome based on the magnitude of the infracentral mean total deviation.

[0040] 9 is a diagram showing an example of an operation screen in the embodiment. The operation screen is displayed, for example, on the output unit 26. A user (e.g., a subject of the determination) operates the input unit 24 (e.g., a keyboard and a mouse) while visually checking the operation screen displayed on the output unit 26, thereby inputting at least the ophthalmic parameters out of the whole-body parameters and the ophthalmic parameters into the determination control unit 251.

[0041] The operation screen has an age input field 201, a gender input field 202, a height input field 203, a weight input field 204, a hypertension input field 205, a diagnosis input field 206, an axial length input field 207, a number of components input field 208, an average total deviation input field 209, and input operation keys 210.

[0042] The average total deviation input field 209 has an input field for the total deviation of the first region 101, "center below TD value 1," an input field for the total deviation of the second region 102, "center below TD value 2," and an input field for the total deviation of the third region 103, "center below TD value 3."

[0043] The age input field 201 is an input field for the subject's age. The gender input field 202 is an input field for the subject's gender. The height input field 203 is an input field for the subject's height. The weight input field 204 is an input field for the subject's weight. The hypertension input field 205 is an input field for whether or not the subject suffers from hypertension. The indication input field 206 is an input field for whether or not the subject's apnea has been indicated by a family member. The axial length input field 207 is an input field for the subject's axial length. The number of ingredients input field 208 is an input field for the number of ingredients in the glaucoma eye drops prescribed to the subject.

[0044] When the input operation key 210 is pressed on the input unit 24 (for example, a mouse or a touch panel), the determination control unit 251 acquires the whole-body parameters and ophthalmologic parameters input using the operation screen and the input unit 24. The determination control unit 251 calculates the subject's body mass index (BMI) based on the subject's height and weight as preprocessing for the determination process. The determination control unit 251 calculates the average value of the value input in the input field "Center Below TD Value 1," the value input in the input field "Center Below TD Value 2," and the value input in the input field "Center Below TD Value 3" as preprocessing for the determination process.

[0045] The determination control unit 251 inputs the acquired whole-body parameters and ophthalmological parameters as explanatory variables to the determination processing unit 252. The determination processing unit 252 uses the input whole-body parameters and ophthalmological parameters and the trained model to perform a determination process (inference process) on the possibility that the subject is suffering from sleep apnea syndrome. The determination control unit 251 or the determination processing unit 252 causes the output unit 26 to display the determination result of the determination processing unit 252.

[0046] Next, an example of the operation of the determination system 1a will be described. Fig. 10 is a flowchart showing an example of the operation of the learning device 10 in this embodiment. In the learning stage of machine learning, the learning control unit 151 acquires teacher data (step S101). The learning processing unit 152 executes a learning process on the teacher data using a machine learning technique. The learning processing unit 152 records a learned model obtained by the learning process in the storage device 12 (step S102).

[0047] 11 is a flowchart showing an example of the operation of the determination device 20 in an embodiment. In an inference stage (determination stage) after the learning stage of machine learning, the determination control unit 251 acquires whole-body parameters and ophthalmic parameters of the subject. The whole-body parameters and ophthalmic parameters of the subject are input from the input unit 14 in response to, for example, a user's operation. The whole-body parameters and ophthalmic parameters of the subject may be acquired via communication from an external communication device (not shown). The whole-body parameters and ophthalmic parameters of the subject may also be acquired from a recording medium (not shown) connected to the determination device 20 (step S201).

[0048] The determination control unit 251 inputs at least the ophthalmic parameters as explanatory variables to the determination processing unit 252 (trained model), causing the determination processing unit 252 to execute a determination process (inference process) (step S202). The determination control unit 251 outputs the determination result of the determination processing unit 252 (output (objective variable) of the trained model) to the output unit 26. The output unit 26 displays the determination result of the determination processing unit 252 (step S203).

[0049] As described above, the storage device 22 stores a trained model obtained based on whole-body parameters (particularly, SAS factors of the whole-body parameters) representing the state of the whole body, ophthalmological parameters (particularly, SAS factors of the ophthalmological parameters) representing the state of the eyes, and a correct label (correct answer information) representing the possibility of the subject suffering from sleep apnea syndrome. The trained model is realized, for example, using a neural network with an extra-tree classifier. The determination unit 25 determines the possibility of the subject suffering from sleep apnea syndrome using the subject's whole-body parameters, the subject's ophthalmological parameters, and the trained model.

[0050] In this way, the determination unit 25 determines the possibility that the subject suffers from sleep apnea syndrome using a trained model obtained based on the whole-body parameters and the ophthalmological parameters. This allows the determination system to more easily determine the possibility that the subject suffers from sleep apnea syndrome. By making a simple determination based on the ophthalmological parameters, it is possible to efficiently connect the subject (patient) to a specialist in sleep apnea syndrome. In addition, it is possible to more accurately determine the possibility that the subject suffers from sleep apnea syndrome.

[0051] (Modification) In the modification of the embodiment, the difference from the embodiment is that the determination result is transmitted from the determination device to a terminal device external to the determination device. In the modification of the embodiment, the difference from the embodiment will be mainly described.

[0052] Fig. 12 is a diagram showing an example configuration of a determination system 1b in a modified example of the embodiment. The determination system 1b is a system that determines the possibility that a subject suffers from sleep apnea syndrome. The determination system 1b includes a learning device 10, a determination device 20, a communication line 30, and a terminal device 40. The determination device 20 is an information processing device such as a server. The determination device 20 includes a communication unit 21, a storage device 22, a storage unit 23, an input unit 24, and a determination unit 25. The determination unit 25 includes a determination control unit 251 and a determination processing unit 252.

[0053] The terminal device 40 is an information processing terminal such as a personal computer, a smartphone terminal, or a tablet terminal. The terminal device 40 is operated by a user. The terminal device 40 includes an input unit 41, a communication unit 42, and an output unit 43. The input unit 41 is a functional unit (operation unit) that accepts operation input, and is, for example, a keyboard, a mouse, or a touch panel. The input unit 41 acquires whole-body parameters and ophthalmologic parameters of the subject in response to operations by the user.

[0054] The communication unit 42 communicates with the communication unit 21 via the communication line 30. The communication unit 42 transmits the whole-body parameters and the ophthalmologic parameters of the subject to the communication unit 21. The communication unit 42 acquires the determination result of the determination processing unit 252 from the communication unit 21.

[0055] The output unit 43 includes a display device such as a liquid crystal display. The output unit 43 displays an operation screen. The output unit 43 displays the determination result of the determination processing unit 252. The output unit 43 may include an audio output unit. The output unit 43 may output the determination result of the determination processing unit 252 by audio.

[0056] Next, an example of the results of cross-validation between a trained model trained using original training data and a trained model re-trained using the original training data and additional training data will be described. Note that the example of the results of cross-validation described below is common to the above-mentioned "embodiment" and "variation of embodiment."

[0057] FIG. 13 is a diagram illustrating an example of an average receiver operating characteristic curve (average ROC curve) with variability for a trained model (hereinafter referred to as the "original trained model") trained using predetermined original training data in an embodiment. The systemic parameters and ophthalmologic parameters in the original training data are based on simple test results for sleep apnea syndrome obtained using a single model of testing equipment. As an example, this single model is the "SAS-2100" manufactured by Nihon Kohden Corporation.

[0058] FIG. 14 illustrates an example of an average receiver operating characteristic curve with variability for a trained model after retraining in an embodiment. The trained model after retraining is a trained model obtained by a retraining process using new training data including the original training data and additional training data. The systemic parameters and ophthalmological parameters in this additional training data are based on, for example, simple test results for 68 people for sleep apnea syndrome obtained using multiple models of testing equipment. For example, these multiple models are the "SAS-2100" from Nihon Kohden Corporation, the "LS-140" from Fukuda Denshi Co., Ltd., and the "Smartwatch PMP-300" from Philips Japan Corporation.

[0059] In the cross-validation results between the original trained model and the retrained trained model, the area under the ROC curve (AUC) of the retrained trained model is equivalent to the area under the ROC curve of the original trained model. Thus, the determination accuracy of the trained model remains almost unchanged regardless of whether a single model of inspection equipment is used or multiple models of inspection equipment are used. In other words, even when the retraining process is performed using training data under conditions different from those of the training data used in the training process of the original trained model (i.e., multiple models of inspection equipment), the determination accuracy of the retrained trained model is maintained.

[0060] Furthermore, the standard deviation of the area under the ROC curve in the trained model after re-training is smaller than the standard deviation of the area under the ROC curve in the original trained model. In this way, the robustness of the trained model is improved by the re-training process using new training data that includes the original training data and additional training data under conditions different from those of the original training data.

[0061] Although an embodiment of the present invention has been described above in detail with reference to the drawings, the specific configuration is not limited to this embodiment, and includes designs within the scope of the gist of the present invention.

[0062] The present invention is applicable to a determination system for determining the possibility that a subject has a predetermined disease.

[0063] 1a, 1b... Determination system, 10... Learning device, 11... Communication unit, 12... Storage device, 13... Storage unit, 14... Input unit, 15... Learning unit, 20... Determination device, 21... Communication unit, 22... Storage device, 23... Storage unit, 24... Input unit, 25... Determination unit, 26... Output unit, 30... Communication line, 40... Terminal device, 41... Input unit, 42... Communication unit, 43... Output unit, 101... First region, 102... Second region area, 103...third area, 151...learning control unit, 152...learning processing unit, 201...age input field, 202...gender input field, 203...height input field, 204...weight input field, 205...hypertension input field, 206...remarks input field, 207...axial length input field, 208...number of components input field, 209...average total deviation input field, 210...input operation keys, 251...determination control unit, 252...determination processing unit

Claims

1. A determination system comprising: a storage device that stores a trained model obtained based on whole-body parameters that represent the condition of the entire body, ophthalmologic parameters that represent the condition of the eyes, and a correct label that represents the possibility of the subject suffering from sleep apnea syndrome; and a determination unit that determines the possibility that the subject suffers from sleep apnea syndrome using the whole-body parameters of the subject, the ophthalmologic parameters of the subject, and the trained model.

2. The determination system of claim 1, wherein the trained model comprises an extra-tree classifier.

3. A determination method executed by a determination system having a storage device that stores a trained model, wherein the trained model is obtained based on whole-body parameters representing the condition of the entire body, ophthalmologic parameters representing the condition of the eyes, and a correct label representing a possibility of the subject suffering from sleep apnea syndrome, and the determination method includes a step of determining the possibility that the subject suffers from sleep apnea syndrome using the whole-body parameters of the subject, the ophthalmologic parameters of the subject, and the trained model.

4. A program for causing a computer to function as a judgment system comprising: a storage device that stores a trained model obtained based on whole-body parameters representing the condition of the entire body, ophthalmologic parameters representing the condition of the eyes, and a correct label representing the possibility of the subject suffering from sleep apnea syndrome; and a judgment unit that uses the whole-body parameters of the subject, the ophthalmologic parameters of the subject, and the trained model to judge the possibility that the subject suffers from sleep apnea syndrome.

Images