Detection cerebrovascular disease in brain images using convolutional neural network

Abstract

Cerebrovascular disease is the world's second major cause of mortality and disability, and the fourth major cause of mortality and disability in South Africa. Cerebrovascular disease occurs due to issues of the brain's blood supply, either the blood supply is cut off or a blood artery within the brain bursts. Radiologists have the responsibility to detect cerebrovascular disease. We now have technology which can help them to better detect this disease. Medical imaging plays an important role in detecting diseases. This study presents implementation of a detection system using artificial intelligence model, namely, convolutional neural networks to help in detecting cerebrovascular disease from magnetic resonance imaging (MRI) scans. Brain images using MRI was obtained from kaggle pub-lic dataset. Segmentation process was applied in this study to normalise images since the images came in different sizes. The effectiveness of the concept was demonstrated using a confusion matrix. The accuracy rate was plotted using the Receiver Operating Characteristics (ROC) curve. The evaluation results show that the Convolutional Neural Network (CNN) model detect cerebrovascular disease successfully with validation accu-racy rate of 90% and test accuracy rate of 80%. The training procedure could be improved by using a larger MRI dataset.