Parkinson's disease progress prediction method based on BCL model

Abstract

The invention discloses a Parkinson's disease progress prediction method based on a BCL model, and relates to the technical field of medical image data, and the method comprises the following steps: S1, obtaining a model comprising Hamp; y score and basic information of the patient, and image data including an MRI image; s2, performing format conversion on the MRI image, and combining with Hamp; y, scoring and carrying out data annotation; s3, preprocessing the image after data annotation; s4, performing dimensionality reduction on the preprocessed image by using principal component analysis, and extracting low-dimensional features of the image; s5, constructing a curriculum learning framework and a classification model based on a balance weight scheduler, and training the dimensionality-reduced features; and S6, predicting the progress of the Parkinson's disease of the patient, and evaluating the prediction performance of each classification model. According to the method, more accurate patient disease stage evaluation can be provided, the calculation efficiency and generalization ability of the model are improved while the data dimension is reduced, and the learning ability of the model for difficult samples is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging data technology, and in particular to a method for predicting Parkinson's disease progression based on a BCL model. Background Art

[0002] Parkinson's disease is a neurodegenerative disease that affects millions of people worldwide. Early diagnosis and monitoring of disease progression are essential for effective treatment. Traditionally, the diagnosis of Parkinson's disease relies mainly on clinical evaluation and observation of patient symptoms. However, this approach may lack accuracy in the early stages of the disease and has difficulty predicting the rate and severity of disease progression. Although some studies have attempted to use machine learning techniques to assist in the diagnosis of Parkinson's disease, most studies have focused on initial diagnosis, and relatively few studies have been conducted on the prediction of disease progression.

[0003] With the rapid development of medical imaging technology, neuroimaging methods such as MRI have been widely used in the study of neurodegenerative diseases. MRI images can provide detailed brain structure information, providing an important tool for the study of Parkinson's disease. However, due to the high data dimension and complex processing of MRI images, how to effectively extract useful features from them and apply them to machine learning models remains a challenge. Therefore, the prediction of Parkinson's disease progression based on MRI images is relatively rare in research.

[0004] In the existing technology, researchers have begun to explore how to combine MRI images with machine learning technology to better perform Parkinson's disease classification tasks. However, most existing studies focus on the binary classification task of Parkinson's disease, lacking in-depth analysis of different stages of disease progression, and in multi-classification tasks, there is still a problem of data imbalance. In addition, existing methods fail to fully utilize the feature differences of MRI images at various stages, resulting in an inability to accurately reflect the severity and progression of the patient's condition. Summary of the invention

[0005] In order to solve the problem that the existing technology cannot accurately predict the progression of Parkinson's disease, the present invention proposes a Parkinson's disease progression prediction method based on the BCL model, which effectively combines MRI images with H&Y scores, and based on unbalanced data, fully utilizes the multi-stage information of MRI images to more comprehensively evaluate the patient's disease progression, so as to solve the above problem.

[0006] The present application discloses a method for predicting Parkinson's disease progression based on the BCL model, comprising the following steps: S1, obtaining clinical data including H&Y scores and basic patient information, and image data including MRI images; S2, convert the format of MRI images and annotate the data in combination with H&Y score; S3, preprocessing the image after data annotation; S4, using principal component analysis to reduce the dimension of the preprocessed image and extract low-dimensional features of the image; S5. Construct a course learning framework based on a balanced weight scheduler, gradually introduce samples of different difficulty levels, construct different classification models and train the features after dimensionality reduction to obtain a course learning model based on a balanced weight scheduler; S6. Use a curriculum learning model based on a balanced weight scheduler to predict the progression of Parkinson's disease in patients and evaluate the prediction performance of each classification model.

[0007] Preferably, S2 comprises the following steps: S21, reading a clinical data file containing an H&Y score and an image data file containing basic information of an MRI image, integrating the two data files of the H&Y score and the basic information of the MRI image into a new field, and formatting and storing the two data files to achieve mapping between the image ID and the H&Y score; S22. Use MRIcroGL to convert MRI images in DICOM format to NIFTI format, and save basic metadata related to the image, including image size, pixel size, image orientation, and time point information; S23. For each NIfTI image, extract the corresponding image ID, and obtain the H&Y score corresponding to this image according to the mapping formed in S21, save the image path of the MRI and the corresponding H&Y score, and realize data labeling of the MRI image.

[0008] Preferably, S3 comprises the following steps: Read the image path and convert the format to float32. For image data that meets the dimension conditions, use OpenCV to resample and normalize the image to the same resolution and size. The dimension conditions are: the image data must be three-dimensional data, and the dimension of each slice must be a valid two-dimensional matrix image data.

[0009] Preferably, the resampling comprises the following steps: Set the target affine matrix size to , resample each slice of each MRI image using the cv2 resize function and return the resampled image object; The standardization process comprises the following steps: Calculate the mean and standard deviation of the pixel values ​​of the input image, subtract the image mean from each pixel value to eliminate the overall brightness effect, and divide it by the sum of the standard deviation and a smoothing term to prevent zero division errors when the standard deviation is zero. Each pixel value after standardization is expressed as the standard deviation number relative to the original data distribution. The calculation formula is as follows:

[0010] in,

[0011]

[0012] is the image pixel value, is the pixel value of the input image, is the image mean, is the image standard deviation, is the smoothing term, is the total number of samples.

[0013] Preferably, S4 comprises the following steps: For all sizes For images of size , flatten the image data into a one-dimensional vector; Use principal component analysis to extract features and achieve data dimensionality reduction. Reintegrate the extracted principal components into the original data frame to obtain a data set, and divide the data set into a training set and a test set, with each principal component as a new feature column.

[0014] Preferably, flattening the image data into a one-dimensional vector comprises the following steps: For each image, straighten its pixel values ​​to form a one-dimensional array, and combine all the one-dimensional arrays into a two-dimensional array with the shape of (number of samples, number of pixels), so as to convert the high-dimensional image into a two-dimensional data matrix; The feature extraction using principal component analysis comprises the following steps: The flattened high-dimensional image data is reduced to 50 dimensions using principal component analysis, which reduces the data dimension while retaining most of the information.

[0015] Preferably, S5 comprises the following steps: S51, divide the training set samples into 5 difficulty levels according to the H&Y score, and divide the training set after dividing the sample difficulty into 4 stages; S52. In the framework of course learning, a balanced weight scheduler is introduced to adjust the weights of samples at different stages, and a course learning framework based on the balanced weight scheduler is obtained; S53, construct classification models including random forest model, support vector machine model, and multi-layer perceptron model; S54, train the model stage by stage. Before the start of each stage, first obtain the number of categories and the number of samples in the training set of the current stage. Then call the balanced weight scheduler to calculate the category weights and generate a weight dictionary. Finally, call the classification models for training respectively, and pass the weight dictionary as the loss function to the classification model. Repeat the above training process until all four stages are trained, and obtain the course learning model based on the balanced weight scheduler.

[0016] Preferably, the four stages of the training set division include: The first stage: only the control group sample with a score of 0 and the easiest sample with a score of 4 were included; Stage 2: Add easier samples with a score of 3 to the first stage; Stage 3: Add more difficult samples with a score of 2 to the second stage; Stage 4: Add the most difficult sample with a score of 1 to the stage 3.

[0017] Preferably, S6 comprises the following steps: The trained curriculum learning model based on the balanced weight scheduler is applied to the test set, and the test samples are classified and predicted. Each sample in the test set is calculated according to the input features, and the output result is the predicted category label for each test sample. The categories include five levels: 0, 1, 2, 3, and 4. The predicted category label corresponds to the H&Y score of the patient's MRI image, which is used to reflect the patient's current stage of Parkinson's disease.

[0018] Preferably, the evaluation indicators for evaluating the predictive performance of each classification model include accuracy, precision, recall, F1 score, mean square error, and specificity.

[0019] Beneficial effects of the present invention: (1) The present invention uses the H&Y score combined with MRI images to predict the progression of Parkinson's disease. Unlike most existing technologies that only focus on the diagnosis of Parkinson's disease, this method can provide a more accurate assessment of the patient's disease stage and help formulate personalized treatment plans.

[0020] (2) The present invention uses PCA dimensionality reduction technology in the traditional machine learning model to extract features, which can effectively reduce the redundant information in MRI images, reduce the data dimension, and improve the computational efficiency and generalization ability of the model, while helping to reduce the risk of model overfitting.

[0021] (3) The present invention introduces a dynamic category weight adjustment mechanism into the framework of course learning. The category weight is dynamically calculated according to the data distribution at each stage, which effectively solves the problem of sample category imbalance, enhances the model's learning ability for difficult samples, and avoids the problem of training being too focused on simple classification samples. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of a method for predicting Parkinson's disease progression based on the BCL model according to an embodiment of the present invention; Figure 2 A flowchart of MRI image preprocessing and feature extraction according to an embodiment of the present invention; Figure 3 A flowchart of constructing a classification model to predict the progression of Parkinson's disease according to an embodiment of the present invention; Figure 4 This is a structural diagram of a course learning framework based on a balanced weight scheduler according to an embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and examples.

[0024] The present application embodiment discloses a method for predicting Parkinson's disease progression based on the BCL model, the process is as follows: Figure 1 As shown, the following steps are included: S1. Obtain the patient's MRI image data and clinical data of H&Y score.

[0025] Clinical data including H&Y scores and basic patient information, as well as image data including MRI images, are obtained from the PPMI platform. In this embodiment, the downloaded image data includes 2781 MRI images of 242 patients, and the downloaded data are stored separately as clinical data files and image data files.

[0026] S2. Convert the format of MRI images and annotate the data based on H&Y scores.

[0027] S21. Read the clinical data file containing the H&Y score, extract the basic information such as the patient number and the consultation time and the H&Y score, and format and store them. Read the image data file containing the basic information of the MRI image, extract the basic information such as the image ID, patient number, gender, age and consultation time, and format and store them. Integrate the two parts of the data file, the H&Y score and the basic information of the MRI image, into a new field by using the patient number and the consultation time, and format and store them, so as to realize the mapping between the image ID and the H&Y score.

[0028] S22. Use MRIcroGL to convert MRI images in DICOM format into NIFTI format, save basic metadata related to the image, including image size, pixel size, image orientation and time point information, and name each MRI image in the form of "image number-MRI imaging information-image acquisition time-image scan sequence number.nii.gz".

[0029] S23. For each NIfTI image, extract the corresponding image ID, and obtain the H&Y score corresponding to this image according to the mapping formed in S21, save the image path of the MRI and the corresponding H&Y score, and realize the data annotation of the MRI image. After the data annotation is completed, there are 2759 MRI images in total, including 1892 images with H&Y score of 0, 195 images with H&Y score of 1, 603 images with H&Y score of 2, 52 images with H&Y score of 3, and 17 images with H&Y score of 4.

[0030] S3, preprocess the images after data annotation, the process is as follows Figure 2 shown.

[0031] Read the image path and convert the format to float32. For image data that meets the dimension conditions, use OpenCV to resample and normalize the image to the same resolution and size. The dimension conditions are: the image data must be three-dimensional data, and the dimension of each slice must be a valid two-dimensional matrix image data.

[0032] Resampling consists of the following steps: Setting the target affine matrix size to , resample each slice of each MRI image using the resize function of cv2, adjust all MRI images to a uniform size, and return the resampled image object.

[0033] The standardization process includes the following steps: calculating the mean and standard deviation of the pixel values ​​of the input image, subtracting the image mean from each pixel value to eliminate the overall brightness effect, and dividing it by the sum of the standard deviation and a smoothing term to prevent zero division errors when the standard deviation is zero. Each pixel value after standardization is expressed as the standard deviation number relative to the original data distribution. The calculation formula is as follows:

[0034] in,

[0035]

[0036] is the image pixel value, is the pixel value of the input image, is the image mean, is the image standard deviation, is the smoothing term, is the total number of samples.

[0037] S4. Use principal component analysis to reduce the dimension of the preprocessed image and extract the low-dimensional features of the image.

[0038] like Figure 2 As shown, for all sizes For images of the same size, the image data is flattened into a one-dimensional vector. Specifically, for each image, its pixel values ​​are straightened to form a one-dimensional array, and all the one-dimensional arrays are combined into a two-dimensional array with the shape of (number of samples, number of pixels), thus converting the high-dimensional image into a two-dimensional data matrix.

[0039] Principal component analysis (PCA) is used for feature extraction to achieve data dimensionality reduction. Specifically, PCA is used to reduce the flattened high-dimensional image data to 50 dimensions, which significantly reduces the data dimension while retaining most of the information. The extracted principal components are reintegrated into the original data frame to obtain a data set, and the data set is divided into a training set and a test set. Each principal component is used as a new feature column as a model input.

[0040] After data preprocessing and feature extraction, the dataset includes 2721 images, including 1886 images with H&Y score of 0, 195 images with H&Y score of 1, 595 images with H&Y score of 2, 30 images with H&Y score of 3, and 15 images with H&Y score of 4. The training set contains 2174 MRI images and the test set contains 547 MRI images, which are saved in the form of pickle files.

[0041] After saving the dataset, load the training and test sets from the pickle file and remove NaN values. Extract 50 features of each MRI image as input features and extract the H&Y score as the output class label.

[0042] S5. Construct a curriculum learning framework based on a balanced weight scheduler, gradually introduce samples of different difficulty levels, construct different classification models and train the features after dimensionality reduction to obtain a curriculum learning model based on a balanced weight scheduler. The flowchart of constructing a classification model to predict the progression of Parkinson's disease is shown in the figure. Figure 3 shown.

[0043] S51. Divide the training set samples into 5 levels of difficulty according to the H&Y score, and divide the training set into 4 stages after dividing the sample difficulty. Since the MRI images of patients with more severe diseases contain more obvious Parkinson's features, the MRI features with a score of 4 (H&Y4) are defined as the easiest samples, the MRI features with a score of 3 (H&Y3) are defined as easier samples, the MRI features with a score of 2 (H&Y2) are defined as more difficult samples, the MRI features with a score of 1 (H&Y1) are the most difficult samples, and the MRI features with a score of 0 are the control group samples. The 4 stages of the training set sample division include: The first stage: only the control group sample with a score of 0 and the easiest sample with a score of 4 were included; Stage 2: Add easier samples with a score of 3 to the first stage; Stage 3: Add more difficult samples with a score of 2 to the second stage; Stage 4: Add the most difficult sample with a score of 1 to the stage 3.

[0044] S52. In the framework of course learning, a balanced weight scheduler is introduced to adjust the weights of samples at different stages, and a course learning framework (BCL) based on a balanced weight scheduler is obtained. A larger weight is given to a smaller number of high-difficulty samples, while a smaller weight is given to a larger number of low-difficulty samples, to alleviate the impact of the imbalance of the number of samples at each stage on the model. The expression is as follows:

[0045] in, represents the weight of the current category, is the total number of samples in the training set, is the number of samples of the current category, is the number of samples for each category, Represents the total number of samples in all categories.

[0046] Specifically, before each stage begins, we first count the samples of different difficulty levels in the current stage, that is, the number of categories of samples in the current stage. Then, we perform weight conversion, calculate the balance weight according to the number of samples in each category and the total number of samples, and pair each category with its corresponding weight. Finally, we convert the pairing result into a weight dictionary and pass it into the subsequent classification model as the weight parameter of the loss function.

[0047] S53. Construct classification models including random forest (RF) model, support vector machine (SVM) model, and multi-layer perceptron (MLP) model.

[0048] For the random forest model, multiple decision trees are constructed, and the final classification result is determined by a voting mechanism. The number of decision trees n_estimators = 50, the maximum depth of each decision tree max_depth = 10, to prevent overfitting caused by a single tree being too complex, the minimum number of samples required for each node to split min_samples_split = 10, the minimum number of samples that must be included in a leaf node min_samples_leaf = 5, and the total number of features considered by the model at each split of each decision tree max_fearures = 'sqrt', that is, the square root of features, introduces additional randomness to prevent all trees from tending to the same structure, and the random seed random_state = 35.

[0049] For the support vector machine model, the classification prediction of different stages of Parkinson's disease is achieved by maximizing the classification boundary. The kernel function uses the Gaussian kernel function to map nonlinear data to high-dimensional space in order to find the optimal classification hyperplane. The expression of the Gaussian kernel function is:

[0050] in, and are the feature vectors of the two input samples, is the Euclidean distance between the two, is a hyperparameter of the kernel function, which controls the decay rate between samples.

[0051] For the multi-layer perceptron model, a feedforward neural network is used to model and implement multi-classification, including two hidden layers. The first layer contains 100 neurons and the second layer contains 50 neurons. The maximum number of training iterations max_iter = 20 is set.

[0052] S54, training the model stage by stage, the course learning framework structure of the balanced weight scheduler in the embodiment of the present application is as follows Figure 4 As shown in the figure, the model is trained stage by stage from the first stage to the fourth stage. Before each stage starts, the number of categories and the number of samples in the current stage training set are obtained; secondly, the balanced weight scheduler is called to calculate the category weights and generate the weight dictionary; finally, the random forest model, support vector machine model, and multi-layer perceptron model are called for training respectively, and the weight dictionary is passed into the classification model as the loss function. The above training process is repeated until all four stages are trained, and a course learning model based on the balanced weight scheduler is obtained.

[0053] S6. Predicting patients’ Parkinson’s disease stage using a curriculum learning model based on a balanced weight scheduler.

[0054] The trained curriculum learning model based on the balanced weight scheduler is applied to the test set, and the test samples are classified and predicted. Each sample in the test set is calculated according to the input features, and the output result is the predicted category label of each test sample, which includes five levels: 0, 1, 2, 3, and 4. The predicted category label corresponds to the H&Y score of the patient's MRI image, which is used to reflect the patient's current stage of Parkinson's disease.

[0055] After the prediction is completed, the performance of each model is evaluated by comparing the predicted results with the true labels; the evaluation indicators used include accuracy, precision, recall, F1 score, mean square error, and specificity. The calculation formula is as follows: Accuracy:

[0056] Accuracy:

[0057] Recall:

[0058] F1 score:

[0059] Mean Square Error:

[0060] Specificity:

[0061] in, is the number of samples correctly predicted as positive, is the number of samples correctly predicted as negative class, is the number of negative samples that are incorrectly predicted as positive. is the number of positive samples that are incorrectly predicted as negative; It is The true value of the samples, It is The predicted value of a sample.

[0062] In a specific embodiment, the performance of different classification models in predicting the progression of Parkinson's disease is compared and analyzed to determine the optimal model in the BCL framework. In this embodiment, the accuracy, precision, recall, F1 score and mean square error of three classification models are compared, and the comparison results are shown in Table 1: Table 1 Comparison of the performance of various classification models in predicting the progression of Parkinson's disease Model Accuracy / % Accuracy / % Recall rate / % F1 Score MSE RF 92.50 92.85 92.50 0.92 0.37 SVM 97.99 98.08 97.99 0.98 0.08 MLP 98.72 98.73 98.72 0.99 0.04 As shown in Table 1, the MLP model performs best in terms of accuracy, mean square error, precision, recall F1 score, and mean square error, proving that in the BCL framework, MLP's prediction is more accurate, the model is more precise, and the accuracy can reach 98.72%. The performance of SVM is second only to MLP, and the overall performance of RF is relatively poor. In general, the accuracy, precision, and recall of the three models in the BCL framework can reach at least 92%, proving that the BCL framework can perform well in multi-classification tasks based on imbalanced samples.

[0063] The specificity comparison of the three classification models in each H&Y score classification is shown in Table 2: Model 0Category 1 Category 2Categories 3 Categories 4 Categories average RF 0.74 1.00 1.00 1.00 1.00 0.95 SVM 0.99 1.00 0.98 1.00 1.00 0.99 MLP 0.98 1.00 0.99 1.00 1.00 0.99 As shown in Table 2, MLP and SVM have similar precision and the lowest misclassification rate, especially in the recognition of category 0. Although it is slightly lower than RF in the recognition of category 2, its overall performance is obviously better. Although RF performs well in other categories, it has more misclassifications in category 0 and its overall performance is slightly worse than the other two models.

[0064] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for predicting Parkinson's disease progression based on the BCL model, characterized in that: The following steps are involved: S1, obtaining clinical data including H&Y scores and basic patient information, and image data including MRI images; S2, convert the format of MRI images and annotate the data in combination with H&Y score; S3, preprocessing the image after data annotation; S4, using principal component analysis to reduce the dimension of the preprocessed image and extract low-dimensional features of the image; S5. Construct a course learning framework based on a balanced weight scheduler, gradually introduce samples of different difficulty levels, construct different classification models and train the features after dimensionality reduction to obtain a course learning model based on a balanced weight scheduler; S6. Use a curriculum learning model based on a balanced weight scheduler to predict the progression of Parkinson's disease in patients and evaluate the prediction performance of each classification model.

2. The method for predicting Parkinson's disease progression based on the BCL model according to claim 1, characterized in that: The S2 comprises the following steps: S21, reading a clinical data file containing an H&Y score and an image data file containing basic information of an MRI image, integrating the two data files of the H&Y score and the basic information of the MRI image into a new field, and formatting and storing the two data files to achieve mapping between the image ID and the H&Y score; S22. Use MRIcroGL to convert MRI images in DICOM format to NIFTI format, and save basic metadata related to the image, including image size, pixel size, image orientation, and time point information; S23. For each NIfTI image, extract the corresponding image ID, and obtain the H&Y score corresponding to this image according to the mapping formed in S21, save the image path of the MRI and the corresponding H&Y score, and realize data labeling of the MRI image.

3. The method for predicting Parkinson's disease progression based on the BCL model according to claim 2, characterized in that: The S3 comprises the following steps: Read the image path and convert the format to float32. For image data that meets the dimension conditions, use OpenCV to resample and normalize the image to the same resolution and size.

4. The method for predicting Parkinson's disease progression based on the BCL model according to claim 3, characterized in that: The resampling comprises the following steps: Set the target affine matrix size to , resample each slice of each MRI image using the cv2 resize function and return the resampled image object; The standardization process comprises the following steps: Calculate the mean and standard deviation of the pixel values ​​of the input image, subtract the image mean from each pixel value to eliminate the overall brightness effect, and divide it by the sum of the standard deviation and a smoothing term to prevent zero division errors when the standard deviation is zero. Each pixel value after standardization is expressed as the standard deviation number relative to the original data distribution. The calculation formula is as follows: in, is the image pixel value, is the pixel value of the input image, is the image mean, is the image standard deviation, is the smoothing term, is the total number of samples.

5. The method for predicting Parkinson's disease progression based on the BCL model according to claim 4, characterized in that: The S4 comprises the following steps: For all sizes For images of size , flatten the image data into a one-dimensional vector; Use principal component analysis to extract features and achieve data dimensionality reduction. Reintegrate the extracted principal components into the original data frame to obtain a data set, and divide the data set into a training set and a test set, with each principal component as a new feature column.

6. The method for predicting Parkinson's disease progression based on the BCL model according to claim 5, characterized in that: The flattening of the image data into a one-dimensional vector comprises the following steps: For each image, straighten its pixel values ​​to form a one-dimensional array, and combine all the one-dimensional arrays into a two-dimensional array with the shape of (number of samples, number of pixels), so as to convert the high-dimensional image into a two-dimensional data matrix; The feature extraction using principal component analysis comprises the following steps: The flattened high-dimensional image data is reduced to 50 dimensions using principal component analysis, which reduces the data dimension while retaining most of the information.

7. The method for predicting Parkinson's disease progression based on the BCL model according to claim 6, characterized in that: The S5 comprises the following steps: S51, divide the training set samples into 5 difficulty levels according to the H&Y score, and divide the training set after dividing the sample difficulty into 4 stages; S52. In the framework of course learning, a balanced weight scheduler is introduced to adjust the weights of samples at different stages, and a course learning framework based on the balanced weight scheduler is obtained; S53, constructing classification models including random forest model, support vector machine model, and multi-layer perceptron model; S54, train the model stage by stage. Before the start of each stage, first obtain the number of categories and the number of samples in the training set of the current stage. Then call the balanced weight scheduler to calculate the category weights and generate a weight dictionary. Finally, call the classification models for training respectively, and pass the weight dictionary as the loss function to the classification model. Repeat the above training process until all four stages are trained, and obtain the course learning model based on the balanced weight scheduler.

8. The method for predicting Parkinson's disease progression based on the BCL model according to claim 7, characterized in that: The four stages of the training set division include: The first stage: only the control group sample with a score of 0 and the easiest sample with a score of 4 were included; Stage 2: Add easier samples with a score of 3 to the first stage; Stage 3: Add more difficult samples with a score of 2 to the second stage; Stage 4: Add the most difficult sample with a score of 1 to the stage 3.

9. The method for predicting Parkinson's disease progression based on the BCL model according to claim 8, characterized in that: The S6 comprises the following steps: The trained curriculum learning model based on the balanced weight scheduler is applied to the test set, and the test samples are classified and predicted. Each sample in the test set is calculated according to the input features, and the output result is the predicted category label for each test sample. The categories include five levels: 0, 1, 2, 3, and 4. The predicted category label corresponds to the H&Y score of the patient's MRI image, which is used to reflect the patient's current stage of Parkinson's disease.

10. The method for predicting Parkinson's disease progression based on the BCL model according to claim 9, characterized in that: The evaluation indicators for evaluating the prediction performance of each classification model include accuracy, precision, recall, F1 score, mean square error, and specificity.

Images