Neurosurgical disease diagnosis and prognosis prediction system and method based on machine learning

Abstract

The present invention relates to the technical field of medical informatization, and in particular to a neurosurgical disease diagnosis and prognosis prediction system and method based on machine learning. The system includes six modules: a data collection module, a data processing module, a disease diagnosis module, a prognosis prediction module, a model evaluation module, and a result output module. The system integrates clinical data, image data and survival data. After preprocessing, screening and feature extraction, support vector machine and random forest algorithms are used to perform disease classification and prognosis prediction. By comparing the results of the algorithms, a diagnosis and a prognosis prediction are obtained. The model evaluation module dynamically adjusts the weights of the algorithms to optimize the prediction accuracy. Finally, the system integrates diagnosis, prediction and evaluation results to generate a visual diagnostic report. This innovative design enables multi-source data fusion, multi-model collaboration and dynamic optimization, significantly improves the diagnostic accuracy and prediction reliability, provides strong support for clinical decision-making, and is expected to play an important role in improving the level of diagnosis and treatment, improving patient prognosis, and optimizing the allocation of medical resources.

Description

Machine Learning-Based Diagnostic and Prognostic Prediction Systems and Methods for Neurosurgical Diseases Technical Field

[0001] This invention relates to the field of medical information technology, and in particular to a system and method for the diagnosis and prognosis prediction of neurosurgical diseases based on machine learning. Background Technology

[0002] With an aging population and changing lifestyles, the incidence of neurosurgical diseases is on the rise, placing enormous pressure on patients' quality of life and social medical resources. Accurate disease diagnosis and prognosis prediction are crucial for developing appropriate treatment plans and improving patient outcomes. Traditional neurosurgical disease diagnosis and prognosis prediction mainly rely on physicians' clinical experience and limited examination methods, which suffer from problems such as high subjectivity, insufficient accuracy, and low efficiency.

[0003] In recent years, with the rapid development of medical imaging technology and artificial intelligence, machine learning-based assisted diagnostic systems have been widely used in the medical field. However, existing neurosurgical disease diagnosis and prognosis prediction systems still have many shortcomings. First, most systems focus only on a single data source, such as using only imaging or clinical data, neglecting the information gain brought by multi-source data fusion. Second, existing systems typically employ a single machine learning model, which is difficult to adapt to the complexity and diversity of neurosurgical diseases. Furthermore, many systems lack dynamic optimization mechanisms, failing to adjust the model in a timely manner based on new data and clinical feedback, leading to a decline in predictive performance over time. In addition, existing systems often present prediction results directly to doctors, lacking expert knowledge intervention and decision support mechanisms, thus limiting the system's application value in actual clinical practice. Summary of the Invention

[0004] In view of the above problems, there is an urgent need to develop a neurosurgical disease diagnosis and prognosis prediction system that can integrate multi-source data, multiple machine learning models, possess dynamic optimization capabilities, and organically combine with expert knowledge. This invention is an innovative solution proposed to address this need.

[0005] This invention proposes a machine learning-based system and method for the diagnosis and prognosis prediction of neurosurgical diseases, including:

[0006] The data acquisition module is used for:

[0007] Collect neurosurgical clinical data, image data, and patient survival data;

[0008] The data processing module, which is communicatively connected to the data acquisition module, is used for:

[0009] Receive neurosurgical clinical data, image data, and patient survival data sent by the data acquisition module;

[0010] The neurosurgical clinical data and patient survival data were preprocessed, screened, and feature-extracted.

[0011] The disease diagnosis module, which is communicatively connected to the data processing module, is used for:

[0012] Based on the data processed by the data processing module, classification prediction is performed using the support vector machine algorithm and the random forest algorithm.

[0013] By comparing the classification prediction results of the support vector machine algorithm and the random forest algorithm, the disease diagnosis result is obtained;

[0014] The prognosis prediction module, which is communicatively connected to the data processing module and the disease diagnosis module, is used for:

[0015] Based on the patient survival data processed by the data processing module, classification and prediction are performed using the support vector machine algorithm and the random forest algorithm.

[0016] The classification prediction results of the support vector machine algorithm and the random forest algorithm are compared to obtain the prognosis prediction result;

[0017] The model evaluation module, which is communicatively connected to the disease diagnosis module and the prognosis prediction module, is used for:

[0018] The prediction results of the disease diagnosis module and the prognosis prediction module are evaluated;

[0019] The weights of the support vector machine algorithm and the random forest algorithm are dynamically adjusted based on the evaluation results.

[0020] The results output module, which is communicatively connected to the disease diagnosis module, prognosis prediction module, and model evaluation module, is used for:

[0021] Receive and integrate the disease diagnosis results, prognosis prediction results, and model evaluation results;

[0022] Generate diagnostic reports and visualizations.

[0023] Preferably, the data processing module includes:

[0024] The data cleaning unit is used to perform outlier and missing value processing on the neurosurgical clinical data and patient survival data.

[0025] The feature selection unit, which is communicatively connected to the data cleaning unit, is used to select highly relevant features from the neurosurgical clinical data and patient survival data.

[0026] The feature extraction unit, communicatively connected to the feature selection unit, is used for:

[0027] Extract high-order clinical features from the neurosurgical clinical data;

[0028] Image features, statistical features, and texture features are extracted from the image data, wherein the texture features include GLCM and Gabor texture features.

[0029] Preferably, the disease diagnosis module includes:

[0030] The Support Vector Machine (SVM) model training unit is used to train the SVM model, learn from the training data, and adjust the model parameters.

[0031] The Random Forest Model Training Unit is used to train the Random Forest Model, learn from the training data, and adjust the model parameters.

[0032] The model fusion unit, which is communicatively connected to the support vector machine model training unit and the random forest model training unit, is used for:

[0033] Receive the prediction results from the support vector machine model and the random forest model;

[0034] Based on a predefined weighting strategy, the prediction results are fused to arrive at the final disease diagnosis result.

[0035] Preferably, the prognosis prediction module includes:

[0036] The survival data processing unit is used to extract features from the patient's survival data, including features such as operation time, postoperative complications, postoperative high fever and physical condition at discharge.

[0037] The prognostic model training unit, which is communicatively connected to the survival data processing unit, is used for:

[0038] Based on the features extracted by the survival data processing unit, a support vector machine model and a random forest model are trained.

[0039] Optimize the parameters of the trained model;

[0040] The prognostic prediction unit, which is communicatively connected to the prognostic model training unit, is used for:

[0041] Use trained support vector machine and random forest models to predict the prognosis of new patient data;

[0042] By combining the prediction results of the two models, the final prognostic prediction result is obtained.

[0043] Preferably, the model evaluation module includes:

[0044] Accuracy evaluation unit, used for:

[0045] The data collected by the data acquisition module is divided into training data and test dataset;

[0046] The accuracy of the classification prediction results was evaluated using a test dataset.

[0047] The model tuning unit, which is communicatively connected to the accuracy evaluation unit, is used for:

[0048] When the accuracy of the test dataset falls below a set threshold, the model is retrained.

[0049] When the accuracy of the test dataset is higher than the set threshold, the current model is confirmed to be usable;

[0050] The weight adjustment unit, which is communicatively connected to the model tuning unit, is used for:

[0051] Based on the accuracy evaluation results, the weights of the support vector machine model and the random forest model in the result fusion are dynamically adjusted.

[0052] As a preferred option, it also includes:

[0053] The data visualization module, which is communicatively connected to the result output module, is used for:

[0054] Receive the diagnostic report generated by the result output module;

[0055] The diagnostic report is transformed into visual charts, including creating visual charts, graphs, heatmaps, and artifact maps.

[0056] Preferably, the data acquisition module includes:

[0057] The clinical data collection unit is used to collect clinical information such as the patient's gender, age, disease course, initial symptoms, preoperative comorbidities, preoperative laboratory results, pathological results, postoperative laboratory results, postoperative complications, and disease outcome.

[0058] The image data acquisition unit is used to acquire the patient's cranial CT and MRI results and convert them into NIFTI format as a neuroimaging atlas.

[0059] The survival data acquisition unit is used to collect survival-related data such as the patient's operation time, whether serious postoperative complications occurred, whether postoperative high fever occurred, and physical condition at discharge.

[0060] As a preferred option, it also includes:

[0061] The expert-assisted decision-making module, which is communicatively connected to the result output module, is used for:

[0062] Receive the diagnostic report and visualization generated by the result output module;

[0063] Provides an interface for doctors to manually review and adjust prediction results;

[0064] Based on doctors' feedback, the instances of manual adjustments were recorded and analyzed to provide a basis for subsequent model optimization.

[0065] As a preferred option, it also includes:

[0066] The cloud deployment module is used for:

[0067] The system is deployed on a cloud server;

[0068] Provide administrators and physicians with different levels of data access and management permissions;

[0069] Enables remote access to the system and synchronization across multiple devices.

[0070] A machine learning-based method for the diagnosis and prognosis prediction of neurosurgical diseases, employing the aforementioned system, includes the following steps:

[0071] S1. Collect patient data, including neurosurgical clinical data, image data, and patient survival data;

[0072] S2. Preprocessing the neurosurgical clinical data and patient survival data, including:

[0073] The K-MEANS clustering algorithm was used to handle outliers and missing values.

[0074] Match the preprocessed clinical data with the image data;

[0075] S3. Perform feature extraction on the preprocessed data, including:

[0076] Extracting high-order clinical features from clinical data;

[0077] Extracting image features, statistical features, and texture features from image data;

[0078] S4. The extracted features are trained using the support vector machine algorithm and the random forest algorithm to obtain a disease diagnosis model;

[0079] S5. Use the aforementioned disease diagnosis model to predict disease diagnoses from new patient data;

[0080] S6. Based on patient survival data, a prognostic prediction model is trained using the support vector machine algorithm and the random forest algorithm;

[0081] S7. Use the prognostic prediction model to predict the prognosis of new patient data;

[0082] S8. Conduct a comprehensive evaluation of disease diagnosis results and prognostic prediction results;

[0083] S9. Based on the evaluation results, dynamically adjust the weights of the support vector machine algorithm and the random forest algorithm in the result fusion;

[0084] S10. Generate diagnostic reports and visualizations, and provide an interface for expert-assisted decision-making.

[0085] The machine learning-based neurosurgical disease diagnosis and prognosis prediction system and method provided by this invention achieves intelligent processing throughout the entire process, from data acquisition, processing, and analysis to result output, through the collaborative work of multiple modules. The innovation and beneficial effects of this system are mainly reflected in the following aspects:

[0086] First, this invention achieves deep fusion of multi-source heterogeneous data. By simultaneously collecting and processing clinical data, imaging data, and patient survival data, the system can comprehensively capture the patient's health status and disease characteristics. This multi-dimensional data integration not only improves the accuracy of diagnosis and prediction but also makes it possible to discover potential disease-related factors.

[0087] Secondly, this invention employs a dual-model collaborative prediction strategy. By combining two machine learning models with different characteristics—support vector machines and random forests—the system can better adapt to the complexity and diversity of neurosurgical diseases. This ensemble learning method not only improves the stability and robustness of predictions but also overcomes the limitations of a single model to some extent.

[0088] Furthermore, this invention introduces a dynamic model optimization mechanism. The system can evaluate model performance in real time based on new data and prediction results, and trigger model retraining when necessary. This adaptive learning capability ensures continuous optimization of system performance, enabling it to adapt to dynamic changes in disease patterns and advancements in medical technology.

[0089] Furthermore, this invention incorporates an expert-assisted decision-making module that organically combines the predictions of artificial intelligence with the professional judgment of doctors. This "AI + expert experience" decision-making model not only improves the interpretability and credibility of the prediction results but also provides clinicians with more comprehensive decision support.

[0090] Finally, the cloud deployment design of this invention significantly improves the system's accessibility and scalability. Through a microservice architecture and load balancing strategy, the system can efficiently handle large-scale concurrent requests, laying the foundation for multi-center, large-sample clinical applications. Meanwhile, strict access control and data encryption mechanisms ensure patient privacy and data security.

[0091] In summary, this invention constructs a comprehensive, accurate, and reliable system for the diagnosis and prognosis prediction of neurosurgical diseases through innovative designs such as multi-source data fusion, multi-model collaboration, dynamic optimization, and expert-assisted decision-making. This system not only significantly improves diagnostic accuracy and predictive reliability but also provides strong support for clinical decision-making, and is expected to play a significant role in enhancing neurosurgical treatment, improving patient outcomes, and optimizing the allocation of medical resources. Attached Figure Description

[0092] Figure 1 is an overall logic block diagram of the system of the present invention;

[0093] Figure 2 is a logic block diagram of the data acquisition module of the present invention;

[0094] Figure 3 is a logic block diagram of the data processing module of the present invention;

[0095] Figure 4 is a logic block diagram of the disease diagnosis module of the present invention;

[0096] Figure 5 is a logic block diagram of the prognosis prediction module of the present invention;

[0097] Figure 6 is a logic block diagram of the model evaluation module of the present invention. Detailed Implementation

[0098] Please refer to Figures 1-6. This invention provides a machine learning-based system and method for the diagnosis and prognosis prediction of neurosurgical diseases. The system includes multiple functional modules, and through data acquisition, processing, analysis, and prediction, it achieves intelligent diagnosis and prognostic assessment of neurosurgical diseases. The invention will now be described in detail with reference to specific embodiments.

[0099] The system of this invention includes a data acquisition module 1, a data processing module 2, a disease diagnosis module 3, a prognosis prediction module 4, a model evaluation module 5, and a result output module 6. These modules are closely connected through data flow and control flow to jointly complete the entire process from data input to result output.

[0100] Data acquisition module 1 is used to collect neurosurgical clinical data, image data, and patient survival data. Preferably, clinical data includes basic patient information (such as age, gender, etc.), symptoms, signs, and laboratory test results. Image data mainly refers to CT and MRI scan results. Patient survival data includes information on surgery, treatment, and follow-up. This module can acquire data through multiple channels such as hospital information systems (HIS) and radiology information systems (RIS), and perform preliminary format standardization and storage.

[0101] Data processing module 2 is communicatively connected to data acquisition module 1, and is used to receive the acquired raw data and perform preprocessing, filtering, and feature extraction. Specifically, this module first cleans the data, including handling missing values ​​and outliers. This invention uses an improved K-means clustering algorithm for outlier detection, the core idea of ​​which is to cluster the data and then identify samples far from the cluster centers. The algorithm is as follows:

[0102] ,

[0103] Let be the objective function. The number of clusters, For the sample size, For the first The class of One sample, For the first Cluster centers of classes.

[0104] For missing values, this invention employs mean / mode imputation or multiple imputation methods depending on the data type. For example, for continuous variables, missing values ​​can be estimated using the following formula:

[0105] ,

[0106] Among them, among them These are the estimated missing values. and They are known and The mean, Is with The set of known observations for observation pairing.

[0107] After data cleaning, data processing module 2 performs feature extraction. For clinical data, it mainly extracts high-order features such as symptom duration and the degree of abnormality in laboratory indicators. For image data, in addition to conventional shape and texture features, this invention also introduces deep learning methods to extract high-level semantic features. For example, a pre-trained convolutional neural network is used to extract features, where the first... The feature map of a layer can be represented as:

[0108] ,

[0109] It is the first Feature map of the layer and These are the weights and biases of the layer, respectively. It is the activation function, and * indicates the convolution operation.

[0110] After feature extraction, data processing module 2 performs feature selection to reduce data dimensionality and improve model efficiency. This invention uses Lasso regression based on L1 regularization for feature selection, with the objective function being:

[0111] ,

[0112] in, For the target variable, For the characteristic matrix, For regression coefficients, This is the regularization parameter. It can be adjusted by... This allows for control over the strictness of feature selection.

[0113] The disease diagnosis module 3 communicates with the data processing module 2, receives the processed data, and performs classification prediction using Support Vector Machine (SVM) and Random Forest (RF) algorithms. The core of SVM is finding the maximum margin hyperplane, and its decision function is:

[0114] ,

[0115] in, For Lagrange multipliers, For category labels, For kernel function, This is a bias term. The present invention employs a radial basis function (RBF) kernel:

[0116] ,

[0117] in, The kernel parameter controls the curvature of the decision boundary.

[0118] Random forests achieve classification by constructing multiple decision trees and voting on them. During the construction of each tree, the criterion for selecting the best splitting feature at node $m$ is to maximize information gain.

[0119] ,

[0120] in, For entropy function, For nodes Data set at the location, As a feature, Features Values A subset of time.

[0121] The disease diagnosis module 3 compares the prediction results of SVM and RF, and uses a weighted voting method to arrive at the final diagnosis. The weights can be dynamically adjusted based on the performance of each model on the validation set.

[0122] This invention's system, through innovations such as multi-source data fusion, multi-model integration, and dynamic weight adjustment, effectively improves the accuracy and reliability of neurosurgical disease diagnosis. The system's modular design also allows for independent optimization and upgrades of each component, providing excellent scalability. In practical applications, this system can serve as an intelligent auxiliary tool for physicians, offering objective and comprehensive diagnostic suggestions, thus contributing to improved diagnostic efficiency and accuracy.

[0123] The present invention includes a prognostic prediction module 4. This module is communicatively connected to the data processing module 2 and the disease diagnosis module 3, and is primarily responsible for predicting the patient's prognosis. Prognostic prediction is a crucial aspect of neurosurgical clinical practice, providing important references for physicians in developing treatment plans and managing patients.

[0124] The prognostic prediction module 4 of this invention employs a machine learning method based on patient survival data. Specifically, this module first extracts features from the patient survival data, focusing on key indicators such as operation time, postoperative complications, postoperative high fever, and physical condition at discharge. These indicators are typically closely related to the prognosis of neurosurgical patients. For example, prolonged operation time may increase the risk of postoperative complications; postoperative high fever may indicate potential infection or inflammatory response; and physical condition at discharge directly reflects the patient's recovery.

[0125] Building upon feature extraction, the prognostic prediction module 4 constructs a prediction model using two algorithms: Support Vector Machine (SVM) and Random Forest (RF). Each algorithm has its advantages in prognostic prediction tasks: SVM excels at handling high-dimensional data and non-linear relationships, while RF effectively handles interactions between features and provides a ranking of feature importance.

[0126] For the SVM model, this invention employs the kernel trick to handle nonlinear relationships. The decision function for prognostic prediction can be expressed as:

[0127] ,

[0128] It is the input feature vector. They are support vectors. These are the corresponding tags. It is a Lagrange multiplier. It is a kernel function used to handle nonlinear relationships. This is a bias term. The present invention preferably uses a radial basis function (RBF) kernel:

[0129] ,

[0130] in, These are kernel parameters that control the complexity of the decision boundary. In practical applications, The value of is usually between 0.01 and 1, and can be optimized through methods such as cross-validation.

[0131] For the random forest model, this invention constructs multiple decision trees, each trained on a randomly selected subset of features. When splitting nodes, the Gini index is used as a measure of impurity.

[0132] ,

[0133] in, For the dataset, For the number of categories, For the first The proportion of class samples. Preferably, the random forest model of the present invention contains 100-500 decision trees, and the maximum depth of each tree is set to the square root of the number of data features, so as to balance the complexity and generalization ability of the model.

[0134] The prognostic prediction module 4 comprehensively considers the prediction results of both the SVM and RF models. This invention uses a weighted average method to fuse the output probabilities of the two models:

[0135] ,

[0136] in, For the final predicted probability, and They are respectively and The predicted probability of the model and These are the corresponding weights. Initially, both weights can be set to 0.5, and can be dynamically adjusted based on the model's performance on the validation set.

[0137] The model evaluation module 5 of this invention is communicatively connected to the disease diagnosis module 3 and the prognosis prediction module 4, and is mainly responsible for evaluating the prediction results and dynamically adjusting the model performance. Model evaluation is a crucial step in ensuring the accuracy and reliability of the system's predictions.

[0138] The model evaluation module 5 of this invention first divides the data collected by the data acquisition module 1 into training data and test datasets. Preferably, a stratified sampling method is used for data division to ensure that the test set can represent the overall data distribution. The proportion of the test set is usually set to 20%-30%, which can be adjusted according to the size of the data.

[0139] For classification tasks (such as disease diagnosis), model evaluation module 5 primarily uses metrics such as accuracy, precision, recall, and F1 score. The formulas for calculating these metrics are as follows:

[0140] Accuracy: ,

[0141] Accuracy: ,

[0142] Recall rate: ,

[0143] F1 score:

[0144] in, For a real example, For a true negative example, As a false positive example, This is a false negative.

[0145] For tasks involving probability estimation, such as prognostic prediction, model evaluation module 5 also calculates metrics such as log loss and AUC-ROC (area under the receiver operating characteristic curve). The formula for calculating log loss is:

[0146] ,

[0147] in, For the sample size, For real labels, To predict probabilities.

[0148] When the accuracy of the test dataset falls below a set threshold, the model evaluation module 5 triggers model retraining. This invention sets different thresholds for different tasks: for disease diagnosis, the accuracy threshold is set to 0.85; for prognosis prediction, the AUC-ROC threshold is set to 0.75. These thresholds are set based on the accuracy requirements in clinical practice, while also considering the feasibility of the model.

[0149] Model evaluation module 5 is also responsible for dynamically adjusting the weights of the support vector machine model and the random forest model in the result fusion. The adjustment strategy is based on the performance of each model over a recent period and can be expressed as:

[0150] ,

[0151] in, For the model The weight, For the model Performance metrics (such as accuracy or AUC-ROC), The temperature parameter controls the smoothness of the weight distribution. A smaller value... This value makes the weight distribution more "sharp," meaning it favors the model with the best performance. In a preferred embodiment of the invention, Set the value to 0.1.

[0152] This invention also includes a data visualization module, which is communicatively connected to the result output module 6. This module can convert diagnostic reports into intuitive visual charts, including but not limited to bar charts, line charts, heatmaps, and pseudo-color charts. For example, for prognostic prediction of cancer patients, Kaplan-Meier survival curves can be used to visually display the changes in survival probability over time for different risk groups. The estimation formula for the survival function is:

[0153] ,

[0154] in, For the observed survival time, For time The number of events, For time The size of the risk set.

[0155] Through these visualization techniques, the system of this invention can more intuitively present diagnostic and predictive results to doctors, helping them to quickly understand key information and improve the efficiency of clinical decision-making.

[0156] The present invention also includes an expert-assisted decision-making module, which is communicatively connected to the result output module 6, and aims to combine the prediction results of artificial intelligence with the professional judgment of doctors to improve the accuracy and reliability of diagnosis and prognosis prediction.

[0157] The expert-assisted decision-making module first receives the diagnostic report and visualization generated by the results output module 6. This information includes the disease diagnosis results, prognostic prediction results, and related statistical data and charts. The module provides an interactive interface that allows doctors to review and adjust the prediction results.

[0158] In practical applications, doctors can use this interface to view the system's predictions and, based on their clinical experience and understanding of the patient's specific situation, confirm, modify, or supplement the results. For example, if the system predicts an 80% probability that a patient has a certain neurological disorder, but the doctor believes, based on other clinical clues, that this probability may be overestimated, they can adjust the predicted probability through the interface.

[0159] Preferably, the expert-assisted decision-making module of the present invention also includes a feedback mechanism. When doctors adjust the prediction results, the system records and analyzes these adjustments. This analysis can help identify potential biases or deficiencies in the model, providing valuable information for subsequent model optimization. For example, if it is found that the system's prediction accuracy in a certain patient group is significantly lower than the overall level, this may indicate the need to collect more training data for this type of patient or adjust the model structure.

[0160] In one embodiment of the present invention, the expert-assisted decision-making module also integrates a case library. This case library contains a large number of historical cases, each including the patient's clinical data, imaging data, diagnostic results, treatment plan, and prognosis. When the system diagnoses and predicts the prognosis of a new patient, the expert-assisted decision-making module automatically retrieves historical cases similar to the current patient from the case library and presents them to the doctor for reference.

[0161] Case similarity is calculated using weighted Euclidean distance:

[0162] ,

[0163] in, and The feature vectors represent the current patient and historical cases, respectively. For the first The weights of each feature, This represents the total number of features. Feature weights can be set through expert experience or determined using machine learning methods such as random forests.

[0164] By referencing similar cases, doctors can more comprehensively assess a patient's current condition and make more accurate diagnoses and prognoses. This integrated decision-making model, combining AI, expert experience, and historical cases, fully leverages the computational power of artificial intelligence and the clinical experience of doctors, and is expected to significantly improve the diagnosis and treatment of neurosurgical diseases.

[0165] This invention relates to a cloud deployment module. This module is designed to improve system accessibility, scalability, and data security. In a preferred embodiment of the invention, the entire neurosurgical disease diagnosis and prognosis prediction system is deployed on a cloud server.

[0166] The cloud deployment module is primarily responsible for deploying the system's various functional modules to the cloud server. This includes data acquisition module 1, data processing module 2, disease diagnosis module 3, prognosis prediction module 4, model evaluation module 5, and results output module 6. By using containerization technology (such as Docker), each module can be encapsulated as an independent container, facilitating deployment and management.

[0167] Preferably, the present invention employs a microservice architecture, which breaks down different functions of the system into multiple independent services. This architecture has the following advantages:

[0168] 1. High scalability: Each service can be scaled independently according to the load.

[0169] 2. Fault isolation: The failure of a single service will not affect the operation of the entire system.

[0170] 3. Technology stack flexibility: Different services can be implemented using the most suitable technology stack.

[0171] The cloud deployment module is also responsible for load balancing the system. When the system faces high concurrency requests, the load balancer can distribute the requests to multiple service instances, improving the system's response speed and stability. This invention employs a load balancing strategy combining Round Robin and Least Connections. The Round Robin strategy can be expressed as:

[0172] ,

[0173] in, The total number of servers. This is the index for the current server.

[0174] The least connections strategy selects the server with the fewest active connections.

[0175] ,

[0176] in, Indicates server The current number of active connections.

[0177] To ensure data security and privacy, the cloud-deployed module implements a strict access control mechanism. The system provides different levels of data access and management permissions for different roles (such as administrators, doctors, researchers, etc.). Access control adopts a Role-Based Access Control (RBAC) model. In the RBAC model, permission allocation can be represented as:

[0178] ,

[0179] in, This represents the set of roles a user owns. This represents the set of permissions that a certain role possesses.

[0180] Furthermore, the cloud-based deployment module enables encrypted data storage and transmission. All sensitive data is encrypted during storage and then encrypted using SSL / TLS protocols during network transmission. This ensures that even in the event of a data breach, unauthorized personnel cannot access patients' private information.

[0181] Finally, this invention also provides a machine learning-based method for the diagnosis and prognosis prediction of neurosurgical diseases. This method closely corresponds to the aforementioned system and includes steps such as data acquisition, data preprocessing, feature extraction, model training, disease diagnosis, prognosis prediction, result evaluation, and visualization.

[0182] During the data acquisition phase, this method simultaneously acquires the patient's clinical data, imaging data, and survival data. Clinical data includes the patient's basic information, symptoms, signs, and laboratory test results; imaging data mainly consists of CT and MRI scan results; and survival data includes surgical details, complications, and follow-up results.

[0183] The data preprocessing step uses the K-MEANS clustering algorithm to handle outliers and impute missing values. The objective function of the K-MEANS algorithm is:

[0184] ,

[0185] in, The number of clusters, For the sample size, For the first The class of One sample, For the first Cluster centers of classes.

[0186] The feature extraction step employs different methods for different types of data. For clinical data, high-order statistical features are primarily extracted; for imaging data, deep learning models are used to extract high-level semantic features. For example, a pre-trained convolutional neural network is used to extract features, where the feature map of layer $l$ can be represented as:

[0187] ,

[0188] in, For the first Layer feature map, and These are the weights and biases of the layer, respectively. For activation function, This indicates a convolution operation.

[0189] In the model training step, this method trains both Support Vector Machine (SVM) and Random Forest (RF) models simultaneously. The decision function of SVM is:

[0190] ,

[0191] in, For Lagrange multipliers, For category labels, For kernel function, This is a bias term.

[0192] Random forests achieve classification by constructing multiple decision trees and using voting. Information gain is used as the evaluation criterion when nodes split.

[0193] ,

[0194] in, It is the entropy function. For nodes The dataset at that location, As a feature, Features Values A subset of time.

[0195] In the disease diagnosis and prognosis prediction steps, this method comprehensively considers the prediction results of SVM and RF, and uses a weighted voting method to obtain the final result. The weights can be dynamically adjusted according to the performance of each model on the validation set.

[0196] The results evaluation process uses multiple metrics, including accuracy, precision, recall, and F1 score. Furthermore, this method implements a dynamic model update mechanism; when model performance drops below a preset threshold, a retraining process is automatically triggered.

[0197] Finally, this method presents the prediction results in an intuitive visualization, including various statistical charts and pseudo-color enhancement of medical images. These visualizations, combined with an expert system, provide physicians with comprehensive and accurate reference information for clinical decision-making.

[0198] Through the above steps, the method of the present invention realizes intelligent diagnosis and accurate prognosis prediction of neurosurgical diseases, providing strong technical support for improving the level of clinical diagnosis and treatment in neurosurgery.

[0199] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A machine learning-based system for the diagnosis and prognosis prediction of neurosurgical diseases, characterized in that, include: The data acquisition module is used for: Collect neurosurgical clinical data, image data, and patient survival data; The data processing module, which is communicatively connected to the data acquisition module, is used for: Receive neurosurgical clinical data, image data, and patient survival data sent by the data acquisition module; The neurosurgical clinical data and patient survival data were preprocessed, screened, and feature-extracted. The disease diagnosis module, which is communicatively connected to the data processing module, is used for: Based on the data processed by the data processing module, classification prediction is performed using the support vector machine algorithm and the random forest algorithm. By comparing the classification prediction results of the support vector machine algorithm and the random forest algorithm, the disease diagnosis result is obtained; The prognosis prediction module, which is communicatively connected to the data processing module and the disease diagnosis module, is used for: Based on the patient survival data processed by the data processing module, classification and prediction are performed using the support vector machine algorithm and the random forest algorithm. The classification prediction results of the support vector machine algorithm and the random forest algorithm are compared to obtain the prognosis prediction result; The model evaluation module, which is communicatively connected to the disease diagnosis module and the prognosis prediction module, is used for: The prediction results of the disease diagnosis module and the prognosis prediction module are evaluated; The weights of the support vector machine algorithm and the random forest algorithm are dynamically adjusted based on the evaluation results. The results output module, which is communicatively connected to the disease diagnosis module, prognosis prediction module, and model evaluation module, is used for: Receive and integrate the disease diagnosis results, prognosis prediction results, and model evaluation results; Generate diagnostic reports and visualizations.

2. The system according to claim 1, characterized in that, The data processing module includes: The data cleaning unit is used to perform outlier and missing value processing on the neurosurgical clinical data and patient survival data. The feature selection unit, which is communicatively connected to the data cleaning unit, is used to select highly relevant features from the neurosurgical clinical data and patient survival data. The feature extraction unit, communicatively connected to the feature selection unit, is used for: Extract high-order clinical features from the neurosurgical clinical data; Image features, statistical features, and texture features are extracted from the image data, wherein the texture features include GLCM and Gabor texture features.

3. The system according to claim 1, characterized in that, The disease diagnosis module includes: The Support Vector Machine (SVM) model training unit is used to train the SVM model, learn from the training data, and adjust the model parameters. The Random Forest Model Training Unit is used to train the Random Forest Model, learn from the training data, and adjust the model parameters. The model fusion unit, which is communicatively connected to the support vector machine model training unit and the random forest model training unit, is used for: Receive the prediction results from the support vector machine model and the random forest model; Based on a predefined weighting strategy, the prediction results are fused to arrive at the final disease diagnosis result.

4. The system according to claim 1, characterized in that, The prognostic prediction module includes: The survival data processing unit is used to extract features from the patient's survival data, including features such as operation time, postoperative complications, postoperative high fever and physical condition at discharge. The prognostic model training unit, which is communicatively connected to the survival data processing unit, is used for: Based on the features extracted by the survival data processing unit, a support vector machine model and a random forest model are trained. Optimize the parameters of the trained model; The prognostic prediction unit, which is communicatively connected to the prognostic model training unit, is used for: Use trained support vector machine and random forest models to predict the prognosis of new patient data; By combining the prediction results of the two models, the final prognostic prediction result is obtained.

5. The system according to claim 1, characterized in that, The model evaluation module includes: Accuracy evaluation unit, used for: The data collected by the data acquisition module is divided into training data and test dataset; The accuracy of the classification prediction results was evaluated using a test dataset. The model tuning unit, which is communicatively connected to the accuracy evaluation unit, is used for: When the accuracy of the test dataset falls below a set threshold, the model is retrained. When the accuracy of the test dataset is higher than the set threshold, the current model is confirmed to be usable; The weight adjustment unit, which is communicatively connected to the model tuning unit, is used for: Based on the accuracy evaluation results, the weights of the support vector machine model and the random forest model in the result fusion are dynamically adjusted.

6. The system according to claim 1, characterized in that, Also includes: The data visualization module, which is communicatively connected to the result output module, is used for: Receive the diagnostic report generated by the result output module; The diagnostic report is transformed into visual charts, including creating visual charts, graphs, heatmaps, and artifact maps.

7. The system according to claim 1, characterized in that, The data acquisition module includes: The clinical data collection unit is used to collect clinical information such as the patient's gender, age, disease course, initial symptoms, preoperative comorbidities, preoperative laboratory results, pathological results, postoperative laboratory results, postoperative complications, and disease outcome. The image data acquisition unit is used to acquire the patient's cranial CT and MRI results and convert them into NIFTI format as a neuroimaging atlas. The survival data acquisition unit is used to collect survival-related data such as the patient's operation time, whether serious postoperative complications occurred, whether postoperative high fever occurred, and physical condition at discharge.

8. The system according to claim 1, characterized in that, Also includes: The expert-assisted decision-making module, which is communicatively connected to the result output module, is used for: Receive the diagnostic report and visualization generated by the result output module; Provides an interface for doctors to manually review and adjust prediction results; Based on doctors' feedback, the instances of manual adjustments were recorded and analyzed to provide a basis for subsequent model optimization.

9. The system according to claim 1, characterized in that, Also includes: The cloud deployment module is used for: The system is deployed on a cloud server; Provide administrators and physicians with different levels of data access and management permissions; Enables remote access to the system and synchronization across multiple devices.

10. A machine learning-based method for the diagnosis and prognosis prediction of neurosurgical diseases, employing the system described in any one of claims 1-9, characterized in that, Includes the following steps: S1. Collect patient data, including neurosurgical clinical data, image data, and patient survival data; S2. Preprocessing the neurosurgical clinical data and patient survival data, including: The K-MEANS clustering algorithm was used to handle outliers and missing values. Match the preprocessed clinical data with the image data; S3. Perform feature extraction on the preprocessed data, including: Extracting high-order clinical features from clinical data; Extracting image features, statistical features, and texture features from image data; S4. The extracted features are trained using the support vector machine algorithm and the random forest algorithm to obtain a disease diagnosis model; S5. Use the aforementioned disease diagnosis model to predict disease diagnoses from new patient data; S6. Based on patient survival data, a prognostic prediction model is trained using the support vector machine algorithm and the random forest algorithm; S7. Use the prognostic prediction model to predict the prognosis of new patient data; S8. Conduct a comprehensive evaluation of disease diagnosis results and prognostic prediction results; S9. Based on the evaluation results, dynamically adjust the weights of the support vector machine algorithm and the random forest algorithm in the result fusion; S10. Generate diagnostic reports and visualizations, and provide an interface for expert-assisted decision-making.

Images