ABSTRACT
Stroke is a serious neurological condition that occurs due to the interruption of blood supply to the brain or rupture of blood vessels, resulting in damage to brain cells. Early identification of stroke is essential for reducing complications and supporting timely medical intervention. This project proposes a Deep Learning-Based Brain Stroke Detection and Explainable Classification System using Computed Tomography (CT) images. The system utilizes the ResNet50 deep learning architecture with transfer learning to classify brain CT images into two categories: NORMAL and STROKE. Image preprocessing techniques such as resizing, grayscale conversion, normalization and data augmentation are applied to improve the quality of input images. Additionally, Explainable Artificial Intelligence (XAI) using SHAP is integrated to identify image regions that contribute to the model's predictions. The trained model is integrated into a Flask-based web application that provides image uploading, classification, confidence analysis and prediction history management using SQLite. The system aims to provide an automated and interpretable approach to brain CT image classification for research and experimental medical image analysis.
INTRODUCTION
Stroke is one of the major neurological disorders affecting people worldwide. It occurs when the blood supply to a part of the brain is interrupted or when a blood vessel ruptures, causing damage to brain cells. Stroke can result in paralysis, speech difficulties, memory problems and other serious neurological complications.
Computed Tomography (CT) imaging plays an important role in the initial examination of patients suspected of having a stroke. It helps identify abnormalities such as bleeding, swelling and structural changes in the brain. However, manually examining CT images requires medical expertise and careful observation. Analysing large numbers of images can also be time-consuming.
Artificial Intelligence and Deep Learning techniques provide new possibilities for automated medical image analysis. Convolutional Neural Networks (CNNs) can automatically extract important visual features from medical images and perform classification.
The proposed project uses the ResNet50 architecture with transfer learning to classify brain CT images into NORMAL and STROKE categories. SHAP-based Explainable AI is incorporated to provide visual explanations of the classification results. A Flask-based web application is developed to allow users to upload CT images and view predictions, confidence scores and explanation visualizations.

Block Diagram