Tdcnnse-Netssd: Three-Dimensional Convolutional Neural Network Based Se-Net Model for Stroke Disease Detection Using Svm Classifier

Abstract

Stroke is among the leading causes of death worldwide and represents a serious medical condition caused by disrupted blood supply to the brain, often leading to long-term disability or fatal outcomes. Early and accurate stroke prediction is therefore critical for timely intervention and reducing mortality rates. In this study, a two-stage architecture is proposed. The first stage employs a hybrid three-dimensional neural network integrated with squeeze-and-excitation blocks to extract meaningful features. In the second stage, classification is carried out using a Support Vector Machine. A total of ten classifiers—including Support Vector Machine, Random Forest, K-Nearest Neighbor, Decision Tree, Naïve Bayes, Voting Classifier, AdaBoost, Gradient Boosting, Multi-Layer Perceptron, and Nearest Centroid—are evaluated and compared with two existing deep learning model such as CNN and LSTM. After applying data balancing through oversampling, ten of these classifiers achieve accuracy levels exceeding 90%. Hyperparameter optimization and cross-validation are applied to all models to further improve performance. Model effectiveness is assessed using accuracy, precision, recall, and F1-score. Experimental results demonstrate that the proposed approach achieves a maximum accuracy of 95.65%, outperforming the other methods.