Improving Brain Stroke Diagnosis by Using Machine Learning Algorithms

Brain stroke is a Cerebrovascular accident that is considered as one of the threatening diseases. Depending on the area of the brain affected and amount of time, the blood supply blockage or bleeding can cause permanent damage or even lead to death. Face to this urgent medical situation, early detection of different stroke aspects can provide significant information that can minimize the stroke attack degree. Nowadays, by the increasing emergence of Artificial Intelligence (AI), Machine Learning (ML) as a branch can successfully aid to reach powerful long term risk predictions. The main contribution of this work is to propose an efficient and accurate ML model that can be used by an expert system of stroke diagnosis, and that can be the most appropriate to correctly predict people at high risk of brain stroke. Our work started by the exploration and the integration of various risk factors related to stroke disease (like: hypertension, heart disease, average of glucose level, body mass index, smoking status…) and data preprocessing. Then, three ML models were developed based on three popular supervised ML algorithms: Support Vector Machine (SVM), Decision Tree (DT) and Artificial Neural Network (ANN). After the models training and testing, the next stage concerned the performance evaluation of the models to better detect and diagnose stroke attack. The most common performance metrics were calculated, namely: accuracy, error rates, sensitivity and specificity. Therefore, the obtained results were encouraging, and compared to the other ML algorithms, ANN model achieved the best accuracy to perfectly predict the stroke.