Comparative Analysis of Brain Stroke Prediction Using Various Pretrained CNN and ViT models

Deep Learning is a technique in which the system analyzes and learns, is one of the most common applications of artificial intelligence that has seen tremendous progress in the digital era. Because of advancements of deep learning in healthcare technology, a growing number of studies have sought to anticipate and assess certain diseases in advance. Stroke illness research is a hot topic right now, especially among the elderly. Stroke, which is lethal in the elderly, is a condition that needs constant medical attention and monitoring due to its high recurrence and fatality rates. So far, the majority of investigations into stroke disease have relied on MRI or CT imaging for straightforward categorization. The goal of this is to use deep learning to detect whether there are initial signs of a brain stroke using CT or MRI images. Using CT or MRI scan pictures, a classifier can predict brain stroke. By using four Pre–trained models such as ResNet-50, Vision Transformer (Vit), MobileNetV2 and VGG-19, we obtained our desired results. The study h ighlights the capabilities of various pretrained CNN models and provides a comparison with Vit models and attempts to discuss limitations of various architectures.