Amongst the most common causes of death globally, stroke is one of top three affecting over 100 million people worldwide annually. There are two classes of stroke, namely ischemic stroke (due to impairment of blood supply, accounting for ~70% of all strokes) and hemorrhagic stroke (due to bleeding), both of which can result, if untreated, in permanently damaged brain tissue. The discovery that the affected brain tissue (i.e., ‘ischemic penumbra’) can be salvaged from permanent damage and the bourgeoning growth in computer aided diagnosis has led to major advances in stroke management. Abiding to the Preferred Reporting Items for Systematic Review and Meta–Analyses (PRISMA) guidelines, we have surveyed a total of 177 research papers published between 2010 and 2021 to highlight the current status and challenges faced by computer aided diagnosis (CAD), machine learning (ML) and deep learning (DL) based techniques for CT and MRI as prime modalities for stroke detection and lesion region segmentation. This work concludes by showcasing the current requirement of this domain, the preferred modality, and prospective research areas.

One of the main causes of death and permanent disability is ischemic stroke, for which prompt and precise diagnosis is essential to successful treatment. This study introduces a novel dual-stream deep learning framework for ischemic stroke classification using Computed Tomography (CT) images, specifically addressing challenges in accuracy, computational efficiency, and clinical interpretability. Three significant innovations are included in the suggested architecture: (1) a hybrid Dual Attention Mechanism that combines Dynamic Routing and Cross-Attention for improved region-specific feature discrimination; (2) a Multi-Scale Feature Extraction Module with parallel convolutional pathways that captures both contextual and fine-grained features; and (3) an Adaptive Random Vector Functional Link layer that significantly reduces training time while maintaining high classification performance. When tested on a single-center dataset, the model achieves state-of-the-art classification accuracy of 98.83% across normal, acute and chronic stroke categories. We demonstrate the strong generalization capabilities of the proposed framework by achieving 92.42% accuracy on a diverse, multi-center dataset of 7,842 CT images. The integration of explainable Artificial Intelligence tools improves clinical trustworthiness by offering clear insight into the model’s decision-making process. These outcomes demonstrate the model’s potential for use in actual clinical settings for quick and accurate stroke diagnosis, along with its interpretability and computational efficiency.