System for predicting probability of depression in parkinson's disease, method for predicting probability of depression in parkinson's disease and program stored in a recording medium

Abstract

According to one embodiment of the present invention, a system for predicting the probability of depression in Parkinson's disease patients may be provided, comprising: a learning model building unit for building a learning model for predicting a risk group for depression in Parkinson's disease patients by machine learning of sample data including data on a plurality of explanatory variables and result variables; a variable selection unit for selecting explanatory variables to be used in a prediction model for predicting the probability of depression in Parkinson's disease patients based on the degree to which each explanatory variable influences the result variable in the built learning model; a prediction model generation unit for generating a prediction model for predicting the probability of depression in Parkinson's disease patients based on the selected explanatory variables; and a prevalence probability calculation unit for calculating the probability of depression in Parkinson's disease patients from data on explanatory variables of the subjects to be predicted based on the generated prediction model.

Description

Technology Field

[0001] The present invention relates to a technology for predicting the probability of depression, and more specifically, to a system, method, and a program stored on a recording medium for predicting the probability of depression in patients with Parkinson's disease. Background Technology

[0002] With the improvement of living standards and the advancement of medical technology, the focus of healthcare is shifting from disease treatment to health promotion and prevention. Consequently, the existing provider-centered medical services are rapidly transitioning into consumer-centered services. As the demand for care increases, there is a growing need for Clinical Decision Support (CDS) in the healthcare sector, which enables safe patient treatment and the pursuit of efficient services. In particular, in primary care settings, medical professionals face various decisions during the course of patient treatment, such as determining which tests to use for patient differentiation and how to proceed with treatment. In this context, CDS refers to support designed to assist medical professionals in making judgments regarding diagnosis and treatment plans. It aims to improve the quality of medical services by applying evidence-based knowledge derived from accumulated clinical data at the point of treatment.

[0003] Meanwhile, in primary care, Parkinson's disease is known as the degenerative disease of old age with the second highest incidence rate after Alzheimer's dementia. In particular, as Korea has the fastest rate of population aging in the world, the prevalence of Parkinson's disease is expected to increase further in the future.

[0004] The symptoms of Parkinson's disease are classified into motor symptoms, such as tremors and rigidity, and non-motor symptoms, such as cognitive impairment. Among the non-motor symptoms, depression occurs most frequently in patients with Parkinson's disease, and it has been reported that one in two patients suffers from depression. Despite the frequent reports of depression in patients with Parkinson's disease, according to the Global Parkinson's Disease Survey Steering Committee (2002), only 1% of patients with Parkinson's disease recognized their depression. The problem to be solved

[0005] Because depressive symptoms resemble not only Parkinson's disease symptoms such as attention deficit, lethargy, and sleep disorders, but also dementia symptoms such as apathy, it is difficult for caregivers of Parkinson's patients to detect depression early. Therefore, early detection and prevention of depression in Parkinson's patients is a critical issue in personalized primary care.

[0006] The present invention was developed to solve the problems of the prior art described above. The objective of the present invention is to provide a system, method, and program for predicting the probability of depression in Parkinson's disease patients that predict the probability of depression in patients with high accuracy by considering complex influencing factors, while presenting the results in an interpretable form that medical professionals can understand. means of solving the problem

[0007] As a means to solve the aforementioned problem, a system for predicting the probability of depression in a Parkinson's disease patient according to one embodiment of the present invention may include: a learning model building unit that builds a learning model for predicting a risk group for depression in a Parkinson's disease patient by machine learning of sample data including data on a plurality of explanatory variables and outcome variables; a variable selection unit that selects explanatory variables to be used in a prediction model for predicting the probability of depression in a Parkinson's disease patient based on the degree to which each explanatory variable influences the outcome variable in the built learning model; a prediction model generation unit that generates a prediction model for predicting the probability of depression in a Parkinson's disease patient based on the selected explanatory variables; and a prevalence probability calculation unit that calculates the probability of depression in a Parkinson's disease patient from data on the explanatory variables of the subject to prediction based on the generated prediction model.

[0008] In one embodiment, a plurality of explanatory variables of the sample data include information regarding sociodemographic factors, health behaviors, environmental factors, medical history, REM sleep behavior disorder, motor symptoms related to Parkinson's disease, and non-motor symptoms related to Parkinson's disease, and the result variable of the sample data may include information regarding the prevalence of depression using the GDS-30 (Geriatric Depression Scale-30).

[0009] In one embodiment, the learning model building unit trains a plurality of stacked ensemble models and calculates model performance statistics for each stacked ensemble model, selects one stacked ensemble model based on the calculated model performance statistics, and the variable selection unit can select explanatory variables based on the degree to which each explanatory variable influences the result variable in the selected stacked ensemble model.

[0010] In one embodiment, the variable selection unit may select explanatory variables used in the prediction model based on the feature importance of the constructed learning model.

[0011] In one embodiment, the plurality of explanatory variables are selected using a stacked ensemble-based prediction algorithm and the corresponding feature importance, and the selected explanatory variables may include (1) Total score of UPDRS, (2) Sum of boxes in CDR, (3) Motor score of UPDRS, (4) Global CDR, (5) K-MoCA, (6) K-MMSE, and (7) REM sleep behavior disorder.

[0012] In one embodiment, the prediction model can calculate the probability of depression in Parkinson's disease patients using a nomogram created based on a Naive Bayesian classifier model.

[0013] In one embodiment, the nomogram may include: a prediction point line representing a prediction score assigned to each explanatory variable and having a score range between 0 and 100; a variable line having a length corresponding to the degree of influence on the probability of prevalence for each explanatory variable and including a start point and an end point that match at least a portion of the score range of the prediction point line; a total point line representing the total sum of the prediction scores calculated for each explanatory variable; and a probability line representing the probability of prevalence corresponding to the total sum of the total point line.

[0014] In one embodiment, the prevalence probability calculation unit calculates the probability of depression in Parkinson's disease patients by reflecting the weights of each explanatory variable, and the weights of each explanatory variable may have larger values ​​in the order of (1) Total score of UPDRS, (2) Sum of boxes in CDR, (3) Motor score of UPDRS, (4) Global CDR, (5) K-MoCA, (6) K-MMSE, and (7) REM sleep behavior disorder.

[0015] In one embodiment, the learning model trained by the learning model building unit may include a stacked ensemble model comprising a plurality of base models that generate base prediction values ​​from multiple risk variables, and a meta model that generates meta prediction values ​​using the prediction values ​​generated from the plurality of base models as training data.

[0016] In one embodiment, the base model may be a Random Forest model, and the meta model may be a logistic regression model that generates meta-predicted values ​​using the predicted values ​​generated from the base model as training data.

[0017] As another means for solving the aforementioned problem, a method for predicting the probability of depression in a Parkinson's disease patient according to one embodiment of the present invention is a method for predicting the probability of depression in a Parkinson's disease patient performed by a system for predicting the probability of depression in a Parkinson's disease patient, and may include: a learning model construction step of constructing a learning model for predicting a risk group for depression in a Parkinson's disease patient by machine learning of sample data including data on a plurality of explanatory variables and outcome variables; a variable selection step of selecting explanatory variables to be used in a prediction model for predicting the probability of depression in a Parkinson's disease patient based on the degree to which each explanatory variable influences the outcome variable in the constructed learning model; a prediction model generation step of generating a prediction model for predicting the probability of depression in a Parkinson's disease patient based on the selected explanatory variables; and a prevalence probability calculation step of calculating the probability of depression in a Parkinson's disease patient from data on the explanatory variables of the subject to be predicted based on the generated prediction model.

[0018] As another means to solve the above-mentioned problem, a program according to one embodiment of the present invention may include a program stored on a recording medium to perform the method for predicting the probability of depression in a Parkinson's disease patient by a computer. Effects of the invention

[0019] According to one embodiment of the present invention, a system for predicting the probability of depression in Parkinson's disease patients, a method for predicting the probability of depression in caregivers, and a program may be provided, which predict the probability of depression in Parkinson's disease patients with high accuracy by considering complex influencing factors and presenting it in an interpretable form that medical professionals can understand, by constructing a learning model for predicting a risk group for depression in Parkinson's disease patients by machine learning of sample data including data on a plurality of explanatory variables and outcome variables, selecting explanatory variables to be used in a prediction model for predicting the probability of depression in Parkinson's disease patients based on the degree to which each explanatory variable influences the outcome variable in the constructed learning model, generating a prediction model for predicting the probability of depression in Parkinson's disease patients based on the selected explanatory variables, and calculating the probability of depression in Parkinson's disease patients from data on explanatory variables of the subject to prediction based on the generated prediction model. Brief explanation of the drawing

[0020] FIG. 1 is a block diagram illustrating the configuration of a system for predicting the probability of depression in a Parkinson's disease patient according to one embodiment of the present invention. FIG. 2 is a diagram showing the configuration of a system for predicting the probability of depression in a Parkinson's disease patient according to one embodiment of the present invention when implemented on a computer. FIG. 3 is a diagram showing the configuration of a system for predicting the probability of depression in a Parkinson's disease patient according to one embodiment of the present invention, implemented by a server and a client terminal. FIG. 4 is a flowchart of a method for predicting the probability of depression in a Parkinson's disease patient implemented by a system for predicting the probability of depression in a Parkinson's disease patient according to one embodiment of the present invention. Figure 5 is a diagram showing the structure of a stacked ensemble model of the present invention. Figure 6 is a graph showing measurement data for REM sleep behavior disorder. Figure 7 is a diagram showing the scatter plot matrix of the motor symptom test and non-motor symptom test results of a Parkinson's disease patient. Figure 8 is a graph comparing the RMSE of machine learning models for predicting depression in Parkinson's disease. Figure 9 is a graph comparing the IA of machine learning models for predicting depression in Parkinson's disease. Figure 10 is a graph comparing the Ev of a machine learning model for predicting depression in Parkinson's disease. FIG. 11 illustrates a nomogram for predicting depression in a Parkinson's disease patient according to one embodiment of the present invention. Specific details for implementing the invention

[0021] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings.

[0022] The embodiments of the present invention are provided to more fully explain the invention to those skilled in the art, and the following embodiments may be modified in various different forms, and the scope of the invention is not limited to the following embodiments. Rather, these embodiments are provided to make the disclosure more faithful and complete and to fully convey the spirit of the invention to those skilled in the art.

[0023] The various embodiments described herein may be implemented, for example, in a recording medium readable by a computer or similar device using software, hardware, or a combination thereof.

[0024] According to hardware implementation, the embodiments described herein may be implemented using at least one of ASICs (application specific integrated circuits), DSPs (digital signal processors), DSPDs (digital signal processing devices), PLDs (programmable logic devices), FPGAs (field programmable gate arrays), processors, controllers, microcontrollers, microprocessors, and electrical units for performing functions.

[0025] According to the software implementation, embodiments such as procedures or functions may be implemented together with separate software modules that perform at least one function or operation. The software code may be implemented by a software application written in a suitable programming language.

[0026] FIG. 1 is a block diagram illustrating the configuration of a depression probability prediction system for a Parkinson's disease patient (10: hereinafter referred to as the 'depression probability prediction system') according to one embodiment of the present invention.

[0027] Referring to FIG. 1, the depression probability prediction system (10) may include a learning model building unit (110), a variable selection unit (130), a prediction model generation unit (150), and a prevalence probability calculation unit (170). The depression probability prediction system (10) may be implemented by a processor and memory installed in an information processing terminal. Additionally, the depression probability prediction system (10) may be implemented by a processing server connected to the information processing terminal via a network. The components of the depression probability prediction system (10) are described in detail below, and these components may be implemented in the form of modules in the information processing terminal or in the form of services provided by a server and a client.

[0028] The learning model building unit (110) can build a learning model to predict the depression risk group of Parkinson's disease patients by machine learning of sample data including data on multiple explanatory variables and result variables.

[0029] Multiple explanatory variables of the sample data may include information regarding sociodemographic factors, health behaviors (e.g., pack-years), environmental factors (e.g., exposure to pesticides), medical history (e.g., diabetes), REM sleep behavior disorder, motor symptoms related to Parkinson's disease (e.g., late motor complications), non-motor symptoms related to Parkinson's disease (e.g., K-MoCA), etc. Figure 5 is a graph illustrating measurement data for REM sleep behavior disorder.

[0030] The outcome variable of the sample data may include information regarding the prevalence of depression. Depression was defined as 'present' or 'absent' using the Geriatric Depression Scale-30 (GDS-30).

[0031] The learning model building unit (110) can build a learning model by learning from sample data that includes information regarding multiple explanatory variables, such as sociodemographic factors, health behaviors, environmental factors, medical history, REM sleep behavior disorder, motor symptoms related to Parkinson's disease, and non-motor symptoms related to Parkinson's disease, through machine learning.

[0032] In one embodiment, the learning model building unit (110) trains a plurality of stacked ensemble models and calculates model performance statistics for each stacked ensemble model, and can select one stacked ensemble model based on the calculated model performance statistics. As described below, in the selected stacked ensemble model, an explanatory variable can be selected by the variable selection unit (130) based on the degree to which each explanatory variable influences the result variable.

[0033] The variable selection unit (130) can select explanatory variables to be used in a prediction model for predicting the probability of depression based on the degree to which each explanatory variable influences the outcome variable in the constructed learning model. The variable selection unit (130) can select explanatory variables to be used in a prediction model based on the feature importance of the variable in the constructed learning model.

[0034] Selected explanatory variables may include K-MMSE, K-MoCA, Global CDR, Sum of boxes in CDR, Total score of UPDRS, Motor score of UPDRS, and REM sleep behavior disorder.

[0035] The prediction model generation unit (150) can generate a prediction model to predict the probability of depression based on selected explanatory variables. The prediction model can calculate the probability of depression using a nomogram created based on a Naive Bayesian classifier model. A nomogram is a diagram that represents the relationship between multiple risk factors in two dimensions to simply and efficiently calculate the predicted probability of diseases, etc.

[0036] A nomogram may include a prediction point line, a variable line, a total point line, and a probability line. The prediction point line represents the prediction score assigned to each explanatory variable and has a score range between 0 and 100. The variable line has a length corresponding to the degree to which it influences the probability of prevalence for each explanatory variable and includes a start point and an end point that match at least a portion of the score range of the prediction point line. The total point line may represent the total sum of the prediction scores calculated for each explanatory variable. The probability line represents the probability of prevalence corresponding to the total sum of the total point line.

[0037] Naive Bayesian classifier models determine the probability for a specific class by applying Bayes' theorem under the assumption that the attribute values ​​or events of the data are independent of each other. Using the assumption of attribute independence, the posterior probability, which is the probability that an individual belongs to a class, is calculated as in [Equation 1].

[0038] [Equation 1]

[0039]

[0040] A nomogram can also be viewed as a model for evaluating how well an individual fits into a specific class. Let c be the target class that the nomogram targets, and When we consider the non-target class as the odds ratio for these two probabilities, it can be expressed as [Equation 2].

[0041] [Equation 2]

[0042]

[0043] [Equation 2] can be expressed as the following [Equation 3].

[0044] [Equation 3]

[0045]

[0046] The prevalence probability calculation unit (170) can calculate the prevalence probability of depression from data on the explanatory variables of the predicted subject based on the generated prediction model.

[0047] The prevalence probability calculation unit (170) can calculate the prevalence probability of depression by reflecting the weights of each explanatory variable. The weights of each explanatory variable may have larger values ​​in the order of (1) Total score of UPDRS, (2) Sum of boxes in CDR, (3) Motor score of UPDRS, (4) Global CDR, (5) K-MoCA, (6) K-MMSE, and (7) REM sleep behavior disorder.

[0048] The learning model trained by the learning model building unit (110) may include a stacked ensemble model comprising a plurality of base models that generate base prediction values ​​from multiple risk variables, and a meta model that generates meta prediction values ​​using the prediction values ​​generated from the plurality of base models as training data. In one embodiment, the stacked ensemble model may be composed of a base model that is a Random Forest model and a meta model that generates meta prediction values ​​using the prediction values ​​generated from the base models as training data.

[0049] Stacked ensemble models create a new model by combining various different machine learning models in a layered manner, and consist of two stages: a base model and a metamodel. The inventors used an artificial neural network (ANN), a random forest, a naive Bayes model, and Classification and Regression Trees (CART) as the base model.

[0050] In the metamodel, the logistic regression (LR) algorithm was applied. The regression algorithm maximizes the stability of the model, increases the confidence value of the base model, and has a low possibility of overfitting to the training data. In the present invention, nine machine learning models (ANN, Random Forest, Naive Bayes, CART, ANN+LR, Random Forest+LR, Naive Bayes+LR, CART+LR, Random Forest+Naive Bayes+CART+ANN+LR) were used as examples. Figure 5 is a diagram showing the structure of the stacked ensemble model of the present invention.

[0052] FIG. 2 is a diagram showing the configuration of a depression probability prediction system (10) according to one embodiment of the present invention when implemented on a computer.

[0053] The computer device (200) in which the depression probability prediction system (10) is implemented refers to an information processing device such as a PC, laptop, smart device, or server. The computer device (100) may include an input device (210), a computation device (220), a storage device (230), and an output device (240).

[0054] The input device (210) can receive multiple sample data and measurements of explanatory variables of the prediction target. The received sample data and measurements of explanatory variables of the prediction target can be stored in the storage device (230).

[0055] The storage device (230) can store input sample data and measurements of explanatory variables of the prediction subject. Additionally, the storage device (230) can store the constructed prediction model and the generated prediction nomogram.

[0056] The computing device (220) can build a prediction model to predict a risk group for depression by learning sample data through machine learning. Additionally, the computing device (220) can select important variables from the built prediction model, generate a prediction nomogram based on the selected important variables, and calculate a predicted value of the probability of depression of the subject using the generated prediction nomogram.

[0057] The output device (240) is a device that outputs the probability of depression in a specific form. The output device (140) may include at least one of a display device, an output device that outputs a document, and a communication device that transmits diabetes prediction information to another device.

[0058] FIG. 3 is a diagram showing the configuration of a depression probability prediction system (10) according to one embodiment of the present invention when implemented by a server (330) and a client terminal (310).

[0059] The depression probability prediction system (10) includes a client terminal (210), a test DB (220), a processing server (230), and a model DB (240).

[0060] The client terminal (210) may be installed in a place such as a public institution, a medical institution, or a patient's home. The client terminal (210) may receive input of sample data or measurements of explanatory variables of the prediction subject. The input sample data or measurements of explanatory variables of the prediction subject may be stored in the inspection DB (220).

[0061] The inspection DB (220) may include a database that stores input sample data and measurements of explanatory variables of the prediction subject.

[0062] The processing server (330) can build a prediction model to predict a group at risk of depression by learning sample data through machine learning. Additionally, the processing server (330) can select important variables from the built prediction model, generate a prediction nomogram based on the selected important variables, and calculate a predicted value of the probability of depression of the subject using the generated prediction nomogram.

[0063] The model DB (340) may contain data related to the constructed prediction model and the generated prediction nomogram. The processing server (330) may store the constructed prediction model in the model DB (340) and the generated prediction nomogram in the model DB (340).

[0064] Although the inspection DB (320), processing server (330), and model DB (340) are shown as separate components in FIG. 3, the inspection DB (320), processing server (330), and model DB (340) can be configured as a single integrated component.

[0065] FIG. 4 is a flowchart of a depression probability prediction method implemented by a depression probability prediction system according to one embodiment of the present invention.

[0066] Referring to FIG. 4, the method for predicting the probability of depression may include a learning model building step (S110), a variable selection step (S130), a prediction model generation step (S150), and a prevalence probability calculation step (S170).

[0067] In the learning model building step (S110), the learning model building unit (110) can build a learning model for predicting a depression risk group by machine learning of sample data including data on multiple explanatory variables and result variables.

[0068] In the variable selection step (S130), the variable selection unit (130) can select explanatory variables to be used in a prediction model for predicting the probability of depression based on the degree to which each explanatory variable influences the result variable in the constructed learning model.

[0069] In the prediction model generation step (S150), the prediction model generation unit (150) can generate a prediction model to predict the probability of depression prevalence based on selected explanatory variables.

[0070] In the disease probability calculation step (S170), the disease probability calculation unit (170) can calculate the probability of depression prevalence from data regarding the explanatory variables of the predicted subject based on the generated prediction model.

[0071] For a detailed description of the learning model construction step (S110), variable selection step (S130), prediction model generation step (S150), and disease probability calculation step (S170), the above description of the learning model construction unit (110), variable selection unit (130), prediction model generation unit (150), and disease probability calculation unit (170) may be referenced.

[0073] The features of the depression probability prediction system and prediction method according to the present invention will be explained below through the description of the development and verification process of the depression probability prediction model.

[0075] [Experimental Data]

[0076] Experimental data was obtained from the epidemiological data of Parkinson's disease from the National Biobank under the Korea Centers for Disease Control and Prevention, and data from 238 Parkinson's disease patients aged 65 or older were analyzed.

[0077] The outcome variable, depression, was defined as 'present' or 'absent' using the Geriatric Depression Scale-30 (GDS-30). The explanatory variables included a total of 35 variables, such as sociodemographic factors, health behaviors (e.g., pack-years), environmental factors (e.g., pesticide exposure), medical history (e.g., diabetes), REM sleep behavior disorder, Parkinson's disease-related motor symptoms (e.g., late motor complications), and Parkinson's disease-related non-motor symptoms (e.g., K-MoCA) included in the PDE data. Figure 6 is a graph showing the measurement data for REM sleep behavior disorder.

[0078] Among the total 343 subjects, 37.6% (129) had depression. The chi-square test results showed significant differences between Parkinson's disease patients without depression and those with depression in REM & RBD, K-MMSE, K-MoCA, Global CDR, Sum of boxes in CDR, Total score of UPDRS, Motor score of UPDRS, K-IADL, H&Y staging, and Chwab & England ADL (p<0.05).

[0079] Figure 7 is a diagram showing the scatter plot matrix of the results of motor symptom tests and non-motor symptom tests of Parkinson's disease patients. By referring to Figure 7, the relationship between variables can be identified from the scatter plot matrix of the results of motor symptom tests and non-motor symptom tests of Parkinson's disease patients.

[0080] [Stacked Ensemble Model]

[0081] In the present invention, an artificial neural network (ANN), a random forest, a naive Bayes model, and Classification and Regression Trees (CART) were used as basic models. A logistic regression (LR) algorithm was applied to the metamodel. Finally, in the present invention, nine machine learning models (ANN, Random Forest, naive Bayes, CART, ANN+LR, Random Forest+LR, naive Bayes+LR, CART+LR, Random Forest+naive Bayes+CART+ANN+LR) were used as examples.

[0082] The predictive performance (e.g., accuracy) of each machine learning model was verified through 10-fold cross-validation. The indices used to evaluate predictive performance were root-mean-square-error (RMSE), index of agreement (IA), and variance of errors (Ev). A lower RMSE indicates higher accuracy of the prediction model, a value closer to 1 indicates greater stability of the model, and a lower Ev indicates greater stability of the model.

[0083] [Selection of Predictors for Depression in Parkinson's Disease]

[0084] The selection of predictor variables was made by (1) comparing the predictive performance of nine machine learning models (ANN, Random Forest, naive Bayes, CART, ANN+LR, Random Forest+LR, naive Bayes+LR, CART+LR, Random Forest+naive Bayes+CART,+ANN+LR) using F-1 scores, and (2) assuming that the model with the lowest RMSE and the highest IA and Ev is the algorithm with the best predictive performance, and selecting it as the final model for predicting Parkinson's depression. The selection of key variables for predicting Parkinson's depression in machine learning algorithms was verified by calculating feature importance using the mean decrease in impurity. In addition, in order to efficiently interpret risk probabilities in the nomogram for predicting the high-risk group for Parkinson's depression in the present invention, only the top 7 variables with high feature importance in the final model were selected to develop the nomogram.

[0085] [Exploring Key Predictors of Depression in Parkinson's Disease Using Basic Models and Stacked Ensemble-Based Machine Learning Algorithms]

[0086] The results regarding the predictive performance (RMSE, IA, Ev) of nine machine learning models for predicting depression in Parkinson's disease are presented in Figures 8 to 10, respectively.

[0087] Figure 8 is a graph comparing the RMSE of machine learning models for predicting depression in Parkinson's disease. The numbers on the x-axis of Figure 8 represent the machine learning models below.

[0088] 1=random forest+LR

[0089] 2=random forest+naive bayes+CART+ANN+LR

[0090] 3 = naive bayes + LR

[0091] 4=ANN+LR

[0092] 5=ANN+LR

[0093] 6=random fores

[0094] 7=naive bayes

[0095] 8=ANN

[0096] 9=CART

[0097] Figure 9 is a graph comparing the IA of machine learning models for predicting depression in Parkinson's disease. The numbers on the x-axis of Figure 9 represent the machine learning models below, respectively.

[0098] 1=random forest+LR

[0099] 2=random forest+naive bayes+CART+ANN+LR

[0100] 3 = naive bayes + LR

[0101] 4=ANN+LR

[0102] 5=ANN+LR

[0103] 6=random fores

[0104] 7=naive bayes

[0105] 8=ANN

[0106] 9=CART

[0108] Figure 10 is a graph comparing the Ev of machine learning models for predicting depression in Parkinson's disease. The numbers on the x-axis of Figure 10 represent the machine learning models below, respectively.

[0109] 1=random forest+LR

[0110] 2=random forest+naive bayes+CART+ANN+LR

[0111] 3 = naive bayes + LR

[0112] 4=ANN+LR

[0113] 5=ANN+LR

[0114] 6=random fores

[0115] 7=naive bayes

[0116] 8=ANN

[0117] 9=CART

[0119] As a result of the analysis, it was confirmed that the stacked ensemble model of Random Forest + Logistic Regression, derived with an RMSE of 0.16, an IA of 0.73, and an Ev of 0.48, had the best predictive performance. Therefore, the inventors determined that the stacked ensemble model of Random Forest + Logistic Regression was the machine learning model with the best performance, and extracted the main variables (characteristics) of Parkinson's disease depression using the stacked ensemble model of Random Forest + Logistic Regression.

[0120] In this study, using a random forest plus logistic stacking ensemble model, the top 7 variables based on feature importance related to Parkinson's disease depression were identified as K-MMSE, K-MoCA, Global CDR, Sum of boxes in CDR, Total score of UPDRS, Motor score of UPDRS, and REM sleep behavior disorder.

[0121] [Development and Validation of a Bayesian Nomogram for Predicting Depression in Parkinson's Disease]

[0122] Figure 11 illustrates a nomogram for predicting depression in Parkinson's disease in Korea using the top 7 variables with high importance in a stacked ensemble model of Random Forest + Logistic Regression.

[0123] The items on the left side of Fig. 11 represent the following items in sequence.

[0124] Test 7-1:UPDRS(Untitled parkinson disease rating) Total UPDRS score

[0125] Test 5-1: CDR (Clinical Dementia Rating) Total Score (sum of boxes)

[0126] Test 8-1:UPDRS(Untitled parkinson disease rating scale) Motor UPDRS score

[0127] Test 4-1: CDR (Clinical Dementia Rating) Global CDR score

[0128] Test 2-1: KMoCA (Korean Montreal Cognitive Assessment) Score

[0129] Test 1-1: KMMSE (Korean mini mental state examination) score

[0130] Presence or absence of REM-1:REM sleep behavior disorders (REM eye movement (REM) sleep behavior disorders; RBD)

[0131] In the developed nomogram, a Parkinson's disease patient with a Total UPDRS score of 105, a CDR total score of 5.5, a Motor UPDRS score of 28, a CDR overall score of 2.5, a KMoCA score of 17, a KMMSE score of 18, and REM sleep behavior disorder was found to have a high-risk predictive probability of Parkinson's disease depression of 89%.

[0133] The steps or processes described above may be executed by hardware components, software components, and / or a combination of hardware components and software components. For example, the steps or processes described in the embodiments may be executed using one or more general-purpose or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are also possible.

[0134] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0135] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0136] Although the embodiments have been described above with reference to limited examples and drawings, those skilled in the art can make various modifications and variations from the description above. For example, suitable results can be achieved even if the described techniques are performed in a different order than described, and / or the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0137] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below.

Claims

Claim 1 A learning model building unit for building a learning model to predict a depression risk group in Parkinson's disease patients by machine learning of sample data including data on multiple explanatory variables and outcome variables; a variable selection unit for selecting explanatory variables to be used in a prediction model to predict the probability of depression in Parkinson's disease patients based on the degree to which each explanatory variable influences the outcome variable in the built learning model; and a prediction model generation unit for generating a prediction model to predict the probability of depression in Parkinson's disease patients based on the selected explanatory variables. The method includes a prevalence probability calculation unit that calculates the probability of depression in Parkinson's disease patients from data regarding explanatory variables of the predicted subject based on the generated prediction model, wherein the plurality of explanatory variables are selected using a stacked ensemble-based prediction algorithm and the corresponding feature importance, and the selected explanatory variables include (1) Total score of UPDRS, (2) Sum of boxes in CDR, (3) Motor score of UPDRS, (4) Global CDR, (5) K-MoCA, (6) K-MMSE, and (7) REM sleep behavior disorder, and the prevalence probability calculation unit calculates the probability of depression in Parkinson's disease patients by reflecting the weights of each explanatory variable, and the weights of each explanatory variable are configured to have larger values ​​in the order of (1) Total score of UPDRS, (2) Sum of boxes in CDR, (3) Motor score of UPDRS, (4) Global CDR, (5) K-MoCA, (6) K-MMSE, and (7) REM sleep behavior disorder, and the prediction model is a Naive Bayesian classifier Calculate the probability of depression in Parkinson's disease patients using a nomogram constructed based on a model, wherein the nomogram represents a prediction point line having a score range between 0 and 100, and the prediction score line represents the prediction score assigned to each explanatory variable;A variable line having a length corresponding to the degree to which it influences the probability of prevalence for each explanatory variable, and including a start point and an end point that match at least a portion of the score range of the prediction score line; a total point line representing the total sum of the prediction scores calculated for each explanatory variable; and a probability line representing the probability of prevalence corresponding to the total sum of the total point line, [Figure]; A system for predicting the probability of depression in patients with Parkinson's disease, wherein the above nomogram is the nomogram described in the above drawing, and among the explanatory variables of the nomogram described in the above drawing, Test 7-1 is the UPDRS (Untitled Parkinson disease rating) Total UPDRS score, Test 5-1 is the CDR (Clinical dementia rating) total score (sum of boxes), Test 8-1 is the UPDRS (Untitled Parkinson disease rating scale) Motor UPDRS score, Test 2-1 is the KMoCA (Korean Montreal Cognitive Assessment) score, Test 1-1 is the KMMSE (Korean Mini Mental State Examination) score, and REM-1 indicates the presence or absence of Rapid Eye Movement (REM) Sleep Behavior Disorders (RBD). Claim 2 A system for predicting the probability of depression in patients with Parkinson's disease according to claim 1, wherein the plurality of explanatory variables of the sample data include information regarding sociodemographic factors, health behaviors, environmental factors, disease history, REM sleep behavior disorder, motor symptoms related to Parkinson's disease, and non-motor symptoms related to Parkinson's disease, and the result variable of the sample data includes information regarding the prevalence of depression using the GDS-30 (Geriatric Depression Scale-30). Claim 3 A system for predicting the probability of depression in Parkinson's disease patients according to claim 1, wherein the learning model building unit trains a plurality of stacked ensemble models and calculates model performance statistics for each stacked ensemble model, selects one stacked ensemble model based on the calculated model performance statistics, and the variable selection unit selects explanatory variables based on the degree to which each explanatory variable influences the result variable in the selected stacked ensemble model. Claim 4 In claim 1, the variable selection unit selects explanatory variables used in the prediction model based on the feature importance of the constructed learning model, in a system for predicting the probability of depression in Parkinson's disease patients. Claim 5 delete Claim 6 delete Claim 7 delete Claim 8 delete Claim 9 A system for predicting the probability of depression in a Parkinson's disease patient according to claim 1, wherein the learning model built by the learning model building unit includes a stacked ensemble model comprising a plurality of base models that generate base prediction values ​​from multiple risk variables, and a meta model that generates meta prediction values ​​using the prediction values ​​generated from the plurality of base models as training data. Claim 10 A system for predicting the probability of depression in Parkinson's disease patients according to claim 9, wherein the base model is a Random Forest model and the meta model is a logistic regression model that generates meta-predicted values ​​using predicted values ​​generated from the base model as training data. Claim 11 A method for predicting the probability of depression in Parkinson's disease patients performed by a system for predicting the probability of depression in Parkinson's disease patients, comprising: a learning model building step for building a learning model to predict a risk group for depression in Parkinson's disease patients by machine learning of sample data including data on multiple explanatory variables and outcome variables; a variable selection step for selecting explanatory variables to be used in a prediction model to predict the probability of depression in Parkinson's disease patients based on the degree to which each explanatory variable influences the outcome variable in the built learning model; and a prediction model generation step for generating a prediction model to predict the probability of depression in Parkinson's disease patients based on the selected explanatory variables.The method includes a prevalence probability calculation step for calculating the probability of depression in Parkinson's disease patients from data regarding explanatory variables of the predicted subject based on the generated prediction model, wherein the plurality of explanatory variables are selected using a stacked ensemble-based prediction algorithm and the corresponding feature importance, and the selected explanatory variables include (1) Total score of UPDRS, (2) Sum of boxes in CDR, (3) Motor score of UPDRS, (4) Global CDR, (5) K-MoCA, (6) K-MMSE, and (7) REM sleep behavior disorder, and in the prevalence probability calculation step, the probability of depression in Parkinson's disease patients is calculated by reflecting the weights of each explanatory variable, and the weights of each explanatory variable are configured to have larger values ​​in the order of (1) Total score of UPDRS, (2) Sum of boxes in CDR, (3) Motor score of UPDRS, (4) Global CDR, (5) K-MoCA, (6) K-MMSE, and (7) REM sleep behavior disorder, and the prediction model is a Naive Bayesian classifier Calculate the probability of depression in Parkinson's disease patients using a nomogram constructed based on a model, wherein the nomogram comprises: a prediction point line representing a prediction score assigned to each explanatory variable and having a score range between 0 and 100; a variable line having a length corresponding to the degree of influence on the probability of prevalence for each explanatory variable and including a start point and an end point that match at least a portion of the score range of the prediction point line; a total point line representing the total sum of the prediction scores calculated for each explanatory variable; and a probability line representing the probability of prevalence corresponding to the total sum of the total point line, [Figure]; A method for predicting the probability of depression in a Parkinson's disease patient, wherein the above nomogram is the nomogram described in the above drawing, and among the explanatory variables of the nomogram described in the above drawing, Test 7-1 is the UPDRS (Untitled Parkinson disease rating) Total UPDRS score, Test 5-1 is the CDR (Clinical dementia rating) total score (sum of boxes), Test 8-1 is the UPDRS (Untitled Parkinson disease rating scale) Motor UPDRS score, Test 2-1 is the KMoCA (Korean Montreal Cognitive Assessment) score, Test 1-1 is the KMMSE (Korean Mini Mental State Examination) score, and REM-1 indicates the presence or absence of Rapid Eye Movement (REM) Sleep Behavior Disorders (RBD). Claim 12 A program stored on a recording medium to perform by a computer the method for predicting the probability of depression in Parkinson's disease patients described in paragraph 11.

Images