Explainable AI for Early Stroke Detection from Brain CT Images

Early and accurate detection of brain stroke is crucial for timely medical intervention and improved patient outcomes. This paper presents an explainable deep learning framework for stroke detection and localization using brain CT images. Present study explores the effectiveness of the deep learning model with the training outcomes of the features. Towards incorporating this objective Explainable AI interface is used. With the CT scan image three differnet model namely convolutional architectures—Light weight CNN, ResNet18, and AlexNet— used to classify ischemic and hemorrhagic strokes. In addition to this Grad-CAM visualizations are employed to highlight discriminative regions within CT scans, enhancing interpretability for clinical decision support. With the help of the experimental work on CT imagery dataset, Resnet 18 architecture found to be providing the high rate of performance accuracy, while Grad-CAM maps effectively localized stroke-affected regions, aligning with radiological patterns. The addition of XAI tools to the deep learning model makes it possible to reduce the gap between the prediction and interpretability, and provides additional support to the radiologist in understanding the outcome based stroke detection model.