Date of Publication :7th October 2025

Abstract: This project proposes a deep learning-based system for automatic brain stroke prediction using CT scan images. Leveraging the Xception model for feature extraction, the system classifies scans as stroke or non-stroke with high accuracy. Comparative evaluation with ResNet50V2 and DenseNet121 models demonstrates Xception’s superior performance, achieving 98.63% accuracy. The system incorporates preprocessing, data augmentation, and transfer learning to enhance robustness and generalization. Designed for real-time diagnosis, it offers a scalable, cost-effective solution suitable for deployment in emergency and low-resource medical settings, supporting faster and more accurate clinical decision-making.

Reference :