Parkinson cognitive disorder classification and prediction model and classification method based on model

Abstract

The Parkinson's cognitive disorder classification prediction model is constructed based on a neural network structure, the neural network structure comprises an input layer, a hidden layer and an output layer, the input layer is used for receiving feature data related to a Parkinson's patient, and the output layer is used for outputting a classification result of the Parkinson's cognitive disorder. The classification method based on the Parkinson cognitive disorder classification and prediction model comprises the following steps: acquiring clinical data of Parkinson and non-Parkinson patients; preprocessing the data; performing feature selection on the processed data, and screening out features related to cognitive impairment; training the data by adopting a neural network, and establishing a cognitive impairment classification prediction model; evaluating the model by using various indexes and optimizing the model; and performing cognitive disorder prediction on a patient to be predicted according to the trained model, and outputting a prediction result. The cognitive impairment of a new patient is effectively predicted, the decision-making process of a clinician is supported, and then the treatment effect and life quality of the patient are improved.

Description

Technical Field

[0001] The present invention relates to the fields of medicine and computer science, and particularly relates to a Parkinson's cognitive impairment classification prediction model and a classification method based on the model. Background Art

[0002] Parkinson's disease (PD) is a common neurodegenerative disease in the middle-aged and elderly population. In addition to common motor symptoms such as resting tremors, increased muscle tone, and bradykinesia, patients with Parkinson's disease often have non-motor symptoms of cognitive impairment, which usually affect five major cognitive domains: long-term memory, attention / working memory, visuospatial ability, executive function, and language. With the continuous aggravation of the population aging trend and the gradual extension of the average life expectancy, the incidence of cognitive impairment shows an increasing trend year by year, which seriously affects the quality of life and survival life of patients, and brings a huge burden to society and families.

[0003] In recent years, artificial intelligence technology (AI) has developed rapidly and shone in various fields. AI has gradually obtained extensive applications in medical data analysis and automated diagnosis research of neuroimages due to its characteristics such as low cost, high accuracy, and stable functions. Machine learning (ML) is the key way to realize AI at present. By exploring rules from past data and training models, it can predict the future, bringing a lot of convenience to people's lives. It mainly includes traditional machine learning and deep learning. There are many traditional machine learning models, such as Classification and Regression Tree (CART), Support Vector Machines (SVM), Naive Bayes, Logistic Regression (LR), K-Nearest Neighbor (KNN), Decision Tree (DT), and so on. Neural network belongs to a special machine learning algorithm, which processes information by simulating the connection form of human brain neurons to carry out learning and prediction. Machine learning has opened up a new path for the diagnosis of Parkinson's disease. It focuses on extracting features from PD patient data, deeply analyzing and processing these data to reveal key information closely related to the occurrence of the disease, so as to achieve accurate analysis and prediction of real events.

[0004] However, it is worth noting that despite the great potential of machine learning, there is still a lack of research on using large-scale and long-term datasets to assist in the early classification prediction of cognitive impairment in PD patients. Summary of the Invention

[0005] To achieve the above object, a Parkinson's cognitive impairment classification prediction model and a classification method based on this model are provided. A variety of indicators are modeled and tested using different combinations, and the efficacy and clinical application value of each model are evaluated to obtain a Parkinson's early cognitive impairment classification prediction model with the best and most stable performance, aiming to provide new methods and ideas for the auxiliary diagnosis of Parkinson's cognitive impairment. The technical solution adopted by the present invention is as follows: A Parkinson's cognitive impairment classification prediction model, the model is constructed based on a neural network structure, the neural network structure includes an input layer, a hidden layer and an output layer, wherein the input layer is used to receive feature data related to Parkinson's patients, and the feature data includes but is not limited to the clinical index data, cognitive assessment data and physiological indexes of the patients. The number of neurons in the input layer is determined according to the dimension of the received feature data, and each neuron has adjustable weights and biases for performing a non-linear transformation on the input data. The output layer is used to output the classification result of Parkinson's patients' cognitive impairment, and the classification result includes healthy people and Parkinson's patients, where Parkinson's patients are classified as having cognitive impairment and not having cognitive impairment.

[0006] A classification method based on a Parkinson's cognitive impairment classification prediction model, the specific steps are as follows: Step 1: Obtain the clinical data of Parkinson's disease patients and non-Parkinson's disease patients; Step 2: Preprocess the data, including but not limited to data cleaning, removing missing values, outliers in the data, and data standardization operations; Step 3: Perform feature selection on the processed data to screen out features related to cognitive impairment; Step 4: Use a neural network to train the data to establish a cognitive impairment classification prediction model for Parkinson's disease patients; Step 5: Evaluate the model using various indicators and optimize the model; Step 6: Predict the cognitive impairment of the patient to be predicted according to the trained model and output the prediction result.

[0007] For further optimization, the data preprocessing in Step 2 is performed using a data preprocessing module, and this data preprocessing module is used to preprocess the feature data related to Parkinson's patients input.

[0008] For further optimization, the feature subset after feature selection is input into the input layer of the model for classification prediction.

[0009] In step 4, the model is trained using a supervised learning algorithm. The training dataset includes relevant feature data of Parkinson's patients with labeled cognitive impairment categories. During the training process, a loss function is used to measure the difference between the model's prediction results and the actual labeled results. The loss function includes, but is not limited to, the cross-entropy loss function. The neuron weights and biases in the model are adjusted through the backpropagation algorithm to minimize the loss function value. When the preset training stop condition is reached, such as reaching a predetermined number of iterations or the loss function value converges below a preset threshold, the model training is stopped.

[0010] For further optimization, in step 5, a model evaluation module is used to evaluate and optimize the model. This model evaluation module is used to evaluate the trained model using an independent test dataset. The test dataset includes relevant feature data of Parkinson's patients with labeled cognitive impairment categories and has not participated in the model training process.

[0011] For further optimization, in step 6, the displayed module is used to display the classification results of Parkinson's patients' cognitive impairment output by the model for the predicted results output.

[0012] For further optimization, it further includes a data storage module for securely storing patient data, intermediate data, and training results during the model training process.

[0013] The beneficial effects of the present invention are as follows: This model uses advanced machine learning algorithms, specifically including three main steps: data preprocessing, feature selection, and model training. In the data preprocessing stage, information including patients' baseline conditions, scale results, blood indicators, etc. is collected and standardized. In the feature selection step, data feature selection is carried out by searching medical literature in recent years. The main databases relied on include, but are not limited to, PubMed, and key features closely related to cognitive impairment are screened out. Finally, by training a neural network model and using external validation to optimize the model performance, the prediction accuracy is improved. The model of the present invention can provide a powerful auxiliary tool for clinicians to help them identify cognitive impairment in Parkinson's patients earlier and more accurately and guide the formulation of personalized treatment plans, thus significantly improving the quality of life of patients. Specific embodiments

[0014] In order to more clearly understand the above-mentioned objects, features, and advantages of the present invention, the present invention will be described in detail below with reference to specific embodiments. The following embodiments are implemented on the premise of the technical solution of the present invention, and detailed implementation manners and specific operation processes are given. However, the present invention can also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the following embodiments.

[0015] A classification prediction model for Parkinson's cognitive impairment, the model is constructed based on a neural network structure, the neural network structure includes an input layer, a hidden layer and an output layer for receiving inputs, the hidden layer receives the input data of the input layer, and the output layer outputs the data of the hidden layer. The input layer is used to receive feature data related to Parkinson's patients, and the feature data includes but is not limited to the clinical index data, cognitive assessment data and physiological indexes of the patients. The number of neurons in the input layer is determined according to the dimension of the received feature data, and each neuron has adjustable weights and biases for performing non-linear transformation on the input data. The output layer is used to output the classification results of Parkinson's patients' cognitive impairment, and the classification results include healthy people and Parkinson's patients, where Parkinson's patients are classified as having cognitive impairment and not having cognitive impairment.

[0016] A classification method based on a classification prediction model for Parkinson's cognitive impairment is as follows: Step 1: Obtain the clinical data of Parkinson's disease patients.

[0017] Step 2: Preprocess the data, including but not limited to data cleaning, removing missing values, outliers, and data standardization operations. The data preprocessing is performed using a data preprocessing module. The preprocessing includes but not limited to data cleaning, removing missing values, outliers, and data standardization operations. First, we use EXCEL for manual screening to remove missing values, outliers, and grouping, and then input the data into the model for data processing, mainly including data normalization, scaling the features to a specific range proportionally while calculating the minimum and maximum values of the features, and performing transformation and standardization on this basis. At the same time, ROSE and SMOTE objects are created for data oversampling to balance the data set.

[0018] Step 3: Perform feature selection on the processed data to screen out features related to cognitive impairment. Feature selection is to select data features by searching medical literature in recent years. The main databases relied on include but are not limited to PubMed. The features include but are not limited to basophils, eosinophils, hemoglobin, lymphocytes, monocytes, neutrophils, platelets, red blood cells, serum chloride, serum glucose, serum potassium, serum sodium, serum uric acid, total protein, white blood cells, age, gender, height, weight, MoCA scale score, and UPDRS I scale score. The feature subset after feature selection is input into the input layer of the model for classification prediction. These variables are tested both individually and in combination to evaluate the performance of the ML method.

[0019] Step 4: Use a neural network to train the data and establish a classification prediction model for cognitive impairment. Download data from the PPMI database, organize and combine the data, and use a neural network to build the model. The model is trained using a supervised learning algorithm. The training dataset includes relevant feature data of Parkinson's patients with marked cognitive impairment categories. During the training process, a loss function is used to measure the difference between the model prediction result and the actual marked result. The loss function includes, but is not limited to, the cross-entropy loss function. The neuron weights and biases in the model are adjusted through the backpropagation algorithm to minimize the loss function value. When the preset training stop condition is reached, such as reaching a predetermined number of iterations or the loss function value converges below a preset threshold, stop the model training.

[0020] Step 5: Evaluate the model using various metrics and optimize the model. The evaluation and optimization of the model are carried out using a model evaluation module. The model evaluation module is used to evaluate the trained model using an independent test data set. The test data set includes relevant feature data of Parkinson's patients with labeled cognitive impairment categories and does not participate in the model training process. The data set is randomly divided into a training set and a test set, with a ratio of 80% for training and 20% for testing, ensuring consistent class distribution. The evaluation metrics include but are not limited to accuracy, AUC (Area Under Curve), recall rate, precision, and F1 score. An ACC curve and a LOSS curve are plotted, a confusion matrix is generated to show the true situation of various predictions, and the area under the ROC curve is calculated. Among them, the accuracy rate is the ratio of the number of correctly predicted samples to the total number of predicted samples, the recall rate is the ratio of the number of predicted positive example samples to the actual number of positive example samples, and the F1 score is the harmonic mean considering both the accuracy rate and the recall rate. The ACC curve intuitively shows the dynamic change of the prediction accuracy of the model on the test set. By carefully observing the trend of the ACC curve, the running efficiency and generalization ability of the model are analyzed. If the curve tends to be stable, it indicates that the model has achieved effective convergence and the training effect is significant. On the contrary, if the curve shows a downward or fluctuating trend, it prompts us to further examine the model, or we may need to adjust the model parameters and expand the training data to optimize the performance. The LOSS curve is another important indicator in the model training process. It records the change trajectory of the loss function value. The loss function is a key indicator for measuring the prediction error of the model. The downward trajectory of the LOSS curve indicates the continuous reduction of the model prediction error and the gradual improvement of the training effect. If the LOSS curve rises against the trend, it may imply potential problems with the model, and we need to re-examine the choice of the loss function or increase the diversity of the training data to obtain a more ideal training effect. All evaluation and adjustment work are carried out under the careful guidance of a professional team to ensure the accuracy of data processing and the scientific nature of model construction. Finally, we choose the value of AUC as the most important evaluation indicator, and the judgment criteria are as follows: AUC = 1 means a perfect classifier; AUC = [0.85, 0.95] means very good results; AUC = [0.7, 0.85] means average results; AUC = [0.5, 0.7] means relatively low results, but it is already very good for prediction; AUC = 0.5 means the same as random guessing, and the model has no predictive value; AUC < 0.5 means worse than random guessing, and predicting the opposite is better than random guessing.

[0021] In addition, the present invention has also undergone external verification. By using an independent data set to evaluate the model on the already established model, the external verification cohort focuses on the performance of the model on new data, which is very beneficial for evaluating the generalization ability of the model. This method highly guarantees that the model can show good performance on new data, thereby improving the accuracy and reliability of prediction.

[0022] Step Six: Predict the cognitive impairment of the patient to be predicted according to the trained model and output the prediction result. The output prediction result is used by the display module to display the classification result of Parkinson's disease patients' cognitive impairment output by the model. The classification results include healthy people and Parkinson's patients, and the Parkinson's patients are further classified into those with cognitive impairment and those without cognitive impairment. This enables medical staff or researchers to understand the classification prediction results made by the model.

[0023] It also includes a data storage module for securely storing patient data, intermediate data, and training results during the model training process. joblib: For persistence (saving objects to disk) for subsequent loading. os: Interacts with the operating system for environment configuration.

[0024] The above shows and describes the main features, usage methods, basic principles, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification only illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will also have various changes and improvements according to actual situations, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. A Parkinson's cognitive impairment classification prediction model, characterized in that: The model is constructed based on a neural network structure, which includes an input layer, a hidden layer and an output layer, wherein the input layer is used to receive feature data related to Parkinson's patients, and the feature data includes but is not limited to the patient's clinical indicator data, cognitive assessment data and physiological indicators. The number of neurons in the input layer is determined according to the dimension of the received feature data, and each neuron has an adjustable weight and bias for performing nonlinear transformation on the input data. The output layer is used to output the classification results of cognitive impairment of Parkinson's patients, and the classification results include healthy people and Parkinson's patients, wherein Parkinson's patients are classified as those with cognitive impairment and those without cognitive impairment.

2. The classification method of the Parkinson's cognitive impairment classification prediction model according to claim 1, characterized in that: The specific steps are as follows: Step 1: Obtain clinical data of Parkinson's disease patients and non-Parkinson's disease patients; Step 2: preprocessing the data, including but not limited to data cleaning, removing missing values ​​and outliers in the data, and data standardization operations; Step 3: Perform feature selection on the processed data to screen out features related to cognitive impairment; Step 4: Using a neural network to train the data to establish a classification prediction model for cognitive impairment in Parkinson's disease patients; Step 5: Use various indicators to evaluate the model and optimize the model; Step 6: Predict cognitive impairment for the patient based on the trained model and output the prediction results.

3. The classification method based on a Parkinson's cognitive impairment classification prediction model as claimed in claim 2, characterized in that: The data preprocessing in step 2 adopts a data preprocessing module, and the data preprocessing module is used to preprocess the input Parkinson's disease patient related characteristic data.

4. The classification method based on a Parkinson's cognitive impairment classification prediction model as claimed in claim 2, characterized in that: The feature subset after feature selection in step three is input into the input layer of the model for classification prediction.

5. The classification method based on a Parkinson's cognitive impairment classification prediction model as claimed in claim 2, characterized in that: In the step 4, the model is trained using a supervised learning algorithm, and the training data set includes relevant feature data of Parkinson's patients with labeled cognitive impairment categories. A loss function is used during the training process to measure the difference between the model prediction results and the actual labeled results. The loss function includes but is not limited to the cross-entropy loss function. The neuron weights and biases in the model are adjusted by the back propagation algorithm to minimize the loss function value. When a preset stopping training condition is reached, such as reaching a predetermined number of iterations or the loss function value converges below a preset threshold, the model training is stopped.

6. The classification method based on a Parkinson's cognitive impairment classification prediction model as claimed in claim 2, characterized in that: In the step five, the model is evaluated and optimized using a model evaluation module, which is used to evaluate the trained model using an independent test data set, wherein the test data set includes relevant feature data of Parkinson's patients with labeled cognitive impairment categories and does not participate in the model training process.

7. The classification method based on a Parkinson's cognitive impairment classification prediction model as claimed in claim 2, characterized in that: The prediction result output in step six is ​​displayed through a display module for the classification result of cognitive impairment of Parkinson's patients output by the model.

8. The classification method based on a Parkinson's cognitive impairment classification prediction model as claimed in claim 2, characterized in that: It also includes a data storage module for securely storing patient data and intermediate data and training results during the model training process.