Edge Computing for Brain Stroke Detection using Deep Learning Techniques

Stroke is a critical medical emergency where every second counts and rapid diagnosis is essential to improving patient outcomes. Traditional diagnostic methods often require sending Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) scan data to central servers, causing delays and risking patient privacy, especially in remote or resource-limited areas. This study presents an edge- computing solution that performs real-time brain stroke detection directly at the point of care using deep learning. By evaluating the accuracies of the Visual Geometry Group 16 (VGG-16) and MobileNet models, we select the best-performing model to develop an edge-based application for predicting brain stroke outcomes. This study mainly focuses on detecting the stroke efficiently and saving time which can be used for the diagnosis process. The dataset that is taken for development of the model are the Computed Tomography (CT) scan images. The images are trained and later validated and tested and sent for model development. The accuracies obtained for VGG-16 and MobileNet are 87.22 and 96.79 percent respectively.