Neural Network-based Prognostic Model for Cerebrovascular Accident using CT Scans

Cerebrovascular accident (CVA), stroke has a 5.5 million yearly mortality rate, making it the second most common cause of death in the world. So, it is inevitable to find a method that can efficiently predict the stroke with high accuracy and considerably in less time. Medical imaging, such as brain MRI or CT scans, helps to provide considerable information on stroke prediction, but a manual prediction can be time-consuming and considerably imprecise. Stroke prediction with empirical attributes can be beneficial but a CT scan provides detailed and accurate images of the brain and helps in the prediction of the stroke in the exact location occurrence. This study employs predicting the stroke by training the CT image dataset using the four efficient algorithms, CNN-EfficientNetB0, Random Forest, Support Vector Machines (SVM), and Decision Trees. With the help of performance measures, the models’ performance is assessed. Our results showed that the early prediction of stroke with the CNN’s EfficientNet-B0 architecture provides high accuracy, sensitivity, and specificity. Hence the proposed method aids in providing an efficient approach for the early prediction of stroke with less time-consuming and comparatively more precise classification.