Brain Stroke Detection Using Deep Learning Algorithms

In the medical field, deep learning techniques attaining popularity in many areas like computer vision (CV), image recognition, natural language processing (NLP), and especially in the domain of medicine. This paper represents to detect the early signs of stroke from MRI using convolutional neural networks. Our methodology, aims to identify MRI images as normal or abnormal cases. In our architecture, several types of CNN, that are ResNET50, alexNET, mobileNET and VGG16 are used. For the task of classification, we input pre-processed images for training the model in the above-mentioned algorithms, and classify normal and abnormal patients. it is supported by flask based environment and mySQL database is connected. The complete back-end system is supported by front-end architecture which includes JS, HTML and CSS. Through our experimental results we concluded that, resNET50 and mobileNET show the best accuracy of 97.8% and 98% respectively among other deep learning models.