A Patient-Centric Smart Healthcare Portal for Brain Disorder Prediction Using MRI/CT Scans and Deep Learning Models

The increasing prevalence of neurological disorders has created a critical need for efficient and precise diagnostic solutions. This study presents an AI-based diagnostic system that automatically detects Alzheimer's disease, brain tumours, and strokes by analyzing MRI/CT scans. The system combines advanced deep learning models with a practical web interface to facilitate early diagnosis and improve clinical workflows. It employs two specialized convolutional neural networks: MobileNetV2, optimized for rapid processing in time-sensitive clinical environments, and InceptionV3, designed for high-precision detection of subtle pathological features. Utilizing transfer learning, both models were trained on MRI datasets with 5-fold cross-validation and SMOTE oversampling to ensure robust performance across disease categories. The implementation includes a Flask-based backend integrated with a user-friendly web platform supporting patient registration, scan uploads, and diagnostic result visualization. Comparative analysis showed that InceptionV3 achieved superior accuracy in identifying complex neurological patterns, while MobileNetV2 provided exceptional speed for routine clinical applications. This integrated framework demonstrates potential as a scalable decision-support tool, combining diagnostic reliability with operational efficiency for adoption in healthcare settings. The system architecture also enables straightforward integration with existing hospital infrastructure and flexibility for future diagnostic expansions.