This project develops a deep learning system to classify brain CT scans as normal or hemorrhagic using MobileNet V2. It processes a dataset of 71,491 DICOM images, achieving ~90% accuracy. The system includes a preprocessing pipeline for efficient data handling and is designed for real-time clinical use.