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

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.

💡 Research Summary

This paper presents a hybrid deep‑learning and machine‑learning framework for the rapid and accurate diagnosis of brain stroke using computed tomography (CT) images. Recognizing that stroke remains a leading cause of death and disability worldwide, the authors address two major shortcomings of existing approaches: reliance on a single slice for prediction and the use of heavyweight convolutional networks that are difficult to deploy in real‑time clinical settings.

To create a more representative dataset, two publicly available CT collections (the Brain Stroke CT Image Dataset with 2,501 images and a second dataset with 2,515 images) were merged, cleaned, and re‑labeled by clinical experts into three categories—Normal, Ischemic, and Hemorrhagic—resulting in a balanced set of 5,016 images.

For feature extraction, five pre‑trained convolutional models—DenseNet201, InceptionV3, MobileNetV2, ResNet50, and Xception—were employed to generate high‑dimensional embeddings from each CT slice. MobileNetV2, noted for its low parameter count and computational efficiency, was highlighted as especially suitable for bedside or edge‑device deployment.

The extracted embeddings were then processed by three dimensionality‑reduction/feature‑enhancement techniques: Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and Bacterial Foraging Optimization (BFO). LDA proved most effective at maximizing inter‑class separability.

Subsequently, seven conventional classifiers—Support Vector Machine (SVC), Random Forest, XGBoost, Decision Tree, Logistic Regression, K‑Nearest Neighbors, and Gaussian Naive Bayes—were trained on each feature‑processing pipeline. The combination of MobileNetV2, LDA, and SVC achieved the highest performance, reaching 97.93 % accuracy, 0.979 F1‑score, and superior precision/recall across all metrics. This result outperformed previously reported accuracies (generally 95 % or lower) while requiring far less computational resources.

The authors discuss limitations, including the modest dataset size and the lack of temporal or multi‑slice context, and suggest future work involving larger multi‑institutional cohorts and sequence‑aware models. Overall, the study demonstrates that a lightweight pre‑trained network coupled with robust feature engineering and a strong classical classifier can deliver state‑of‑the‑art stroke detection from CT images, offering a practical pathway toward real‑time, cost‑effective clinical decision support.