Machine Learning based Approach for Predicting Brain Stroke from Electronic Health Records

The aim of the proposed work is to enhance the early diagnosis rate of brain stroke so that the number of deaths and long-term damage cases can be minimized through timely medical treatment. The given system was designed using the Django framework, which provides an interface for user authentication, outcome prediction, and uploading images. Confusion matrices are used to test a variety of models with the aim that their outcomes are accurate. The data is stored in a MySQL database. As stroke ranks within the world’s top causes of mortality and disability, it is important to develop reliable and efficient diagnostic technologies. Traditional stroke detection approaches involve radiologists manually interpreting computed tomography(CT) scan pictures, which can be time-consuming and error-prone. In order to automate the procedure, this research makes use of machine learning algorithms that can swiftly and reliably evaluate CT scan pictures. The GLCM approach extracts textural information from images, allowing the system to distinguish between healthy and stroke-affected brain tissues. Random Forest, KNN, and SVM machine learning algorithms are used in the study. The application allows users to submit a CT scan image, and get predictions about the risk of having a stroke.