An Explainable AI for Detection and Classification of Microwave Brain Stroke

Stroke is a serious neurological condition that requires timely diagnosis and localization for appropriate therapy. Microwave imaging provides a promising non-invasive and portable alternative for head stroke localization and classification. This work proposes a method to solve this complex problem by presenting a wavelet convolutional neural network (CNN), which combines multiresolution analysis with CNN to learn distinctive patterns in the scalogram for accurate classification. An explainable artificial intelligence approach is proposed to explain the model and to point out distinguishing properties for classifying stroke types. Grad-CAM visualization is utilized to enhance interpretability and boost physician confidence. The experimental analysis shows an improvement of localization accuracy and sensitivity.