A Novel Approach for Early Brain Stroke Detection Using Genetic Algorithm

Stroke is a main source of mortality, disability in the world, and early detection is the only way for effective intervention and treatment of it. This research, we have compared methods by using ML and DL for early diagnosis of stroke using brain imaging data. While CNN, ResNet, and BiLSTM are used in DL, SVM and KNN algorithms are used ML with the genetic algorithm. It categorizes images of stroke and normal brain to provide accurate analysis. Each preprocessing step has been undertaken for ensuring quality data consistency. In the machine learning field, the experiment shows at an early detection stage the deep learning techniques are better than that of conventional methods. Thus, they can automatically define features more prominently, recognize complex patterns that are likely to be present within medical images due to models of DL, which further aids in stroke detection. Performance metrics of the each algorithm will be tested and the best performing model is identified when a stroke is found. CNN showed higher accuracy and performance evaluated with a accuracy 99% followed by BiLSTM at 90%, while traditional ML models had much lower values, with KNN at 88%, SVM at 86%, and ResNet at 86% for these parameters. In fact, these approaches are going to give a direction on whether these methods might be taken to the clinic and door-opening instruments for better early diagnosis tools of strokes.