An Efficient Deep Learning Framework for Brain Stroke Diagnosis Using Computed Tomography Images

TL;DR

The study tackles rapid, accurate three-class brain stroke diagnosis from CT images by extracting features with multiple pre-trained CNNs and classifying them via traditional ML models after applying feature optimization. The standout configuration—MobileNetV2 with Linear Discriminant Analysis for dimensionality reduction and Support Vector Classifier for classification—achieves 97.93% accuracy with high precision and recall, while maintaining computational efficiency. Across extensive 10-fold cross-validation and comprehensive comparisons to other architectures and optimization strategies, the approach demonstrates robust performance and practical viability for real-time clinical use. The work also outlines limitations and future directions, including self-supervised learning, transformer-based models, and multi-modal data integration to further enhance generalizability and diagnostic accuracy.

Abstract

Brain stroke is a leading cause of mortality and long-term disability worldwide, underscoring the need for precise and rapid prediction techniques. Computed Tomography (CT) scan is considered one of the most effective methods for diagnosing brain strokes. Most stroke classification techniques use a single slice-level prediction mechanism, requiring radiologists to manually select the most critical CT slice from the original CT volume. Although clinical evaluations are often used in traditional diagnostic procedures, machine learning (ML) has opened up new avenues for improving stroke diagnosis. To supplement traditional diagnostic techniques, this study investigates machine learning models for early brain stroke prediction using CT scan images. This research proposes a novel machine learning approach to brain stroke detection, focusing on optimizing classification performance with pre-trained deep learning models and advanced optimization strategies. Pre-trained models, including DenseNet201, InceptionV3, MobileNetV2, ResNet50, and Xception, are used for feature extraction. Feature engineering techniques, including BFO, PCA, and LDA, further enhance model performance. These features are then classified using machine learning algorithms, including SVC, RF, XGB, DT, LR, KNN, and GNB. Our experiments demonstrate that the combination of MobileNetV2, LDA, and SVC achieved the highest classification accuracy of 97.93%, significantly outperforming other model-optimizer-classifier combinations. The results underline the effectiveness of integrating lightweight pre-trained models with robust optimization and classification techniques for brain stroke diagnosis.

Figures (7)

• Figure 1: Proposed Architecture for Brain Stroke Detection
• Figure 2: Sample images of Computed Tomography (CT) scan in normal, hemorrhagic, and Ischemic states.
• Figure 3: Bar Charts compare the classification performance of different feature selection techniques applied with MobileNetV2 and different classifiers: (a) No Feature Selection, (b) Bacterial Foraging Optimization (BFO), (c) Principal Component Analysis (PCA), and (d) Linear Discriminant Analysis (LDA). Each feature selection technique shows the comparison of accuracy across seven different classifiers.
• Figure 4: Learning curves comparing the performance of different feature optimization techniques employed with MobileNetV2 in terms of accuracy score with varying training set sizes. (a) Extreme Gradient Boosting with No Feature Optimization, (b) Bacterial Foraging Optimization with k-Nearest Neighbors, (c) Principal Component Analysis with Support Vector Classifier, (d) Linear Discriminant Analysis and Support Vector Classifier.
• Figure 5: 10 Fold Cross Validation For MobileNetV2+LDA+SVC