HemoScan: Machine Learning-based Brain Hemorrhage and Tumor Diagnosis using MATLAB Image Processing and Gradient Boosting Models

The early detection of brain hemorrhages and tumors is vital for improving patient outcomes and providing timely medical help. Traditional methods that depend on manual interpretation of MRI scans take a lot of time and are often affected by human error and inconsistency. To solve these problems, this study introduces HemoScan, an automated diagnostic system that combines MATLAB-based image preprocessing with machine learning classifiers for effective brain abnormality detection. The system uses Anisotropic Diffusion Filtering to reduce noise, Histogram Equalization to improve contrast, and Morphological Operations for better skull stripping. This process creates clear and consistent images for analysis. It also extracts important image features, including intensity, texture, and shape. These features train two advanced gradient boosting models. XGBoost achieves a hemorrhage detection accuracy of 96.5%, while LightGBM reaches 99.95% accuracy in distinguishing between benign and malignant tumors. The models are assessed using accuracy, sensitivity, specificity, and F1-score, showing high reliability and robustness. To ensure it works in real settings, HemoScan offers a user-friendly web interface made with Flask. This allows clinicians to upload MRI scans and receive diagnostic feedback in real time. Overall, this system provides a practical, lightweight, and understandable solution suitable for both well- equipped hospitals and areas with limited resources.