ARCHIVES

Brain stroke is a critical neurological emergency, often causing long-term disability or mortality if not diagnosed in time. This research presents a practical implementation of an AI-based framework using Convolutional Neural Networks (CNNs) for detecting brain strokes from CT images and predicting severity levels. By leveraging deep learning and labeled neuroimaging datasets, our model demonstrates early and accurate classification of stroke versus non-stroke conditions, along with severity estimation. The model was trained on a curated dataset with structured preprocessing, and the pipeline includes performance metrics for reliability. We further address key considerations like model generalizability, data governance, and explain ability. The paper contributes both a replicable codebase and an empirical foundation for clinical AI deployment.