Detection of Brain Stroke Disease Using Deep Learning Techniques

Early stroke detection is essential for effective treatment and prevention of long-term disability. Deep learning methods have shown promising results in detecting various medical conditions, including stroke. Through this study, a strategy for identifying brain stroke disease using deep learning techniques and image preprocessing is provided. The methodology involves collecting a diverse and balanced dataset of brain scans, preprocessing the data to extract relevant features, training a deep learning model, tuning hyperparameters, and evaluating the model. Two transfer learning models have been used, MobileNet and ResNet. Among them, MobileNetV3Small has the highest validation accuracy which is 88%. For ResNet50 the overall accuracy is 99% and for MobileNetV3Small it is 96%. The proposed methodology can potentially improve the accuracy and efficiency of stroke diagnosis and enable early intervention, leading to better patient outcomes.