A narrative review on collateral circulation classification for ischemic stroke

This review explores recent advancements in the classification of collateral circulation in ischemic stroke, with a focus on artificial intelligence-driven imaging analysis. It provides a comprehensive overview of stroke types, the physiological importance of collateral networks, and compares imaging modalities such as CT, MRI, and Cone Beam Computed Tomography (CBCT) for collateral assessment. The paper also examines various collateral scoring systems, highlighting both manual and automated methods. Particular attention is given to machine learning and deep learning classification techniques, including convolutional neural networks, residual networks, and multimodal fusion strategies aimed at improving diagnostic accuracy and reproducibility. Key challenges such as data variability, model generalizability, and explainability are discussed, along with future research directions to enhance clinical applicability. This review underscores the increasing role of AI-based classification in enabling faster, more accurate collateral evaluation and improving patient outcomes in ischemic stroke management. • Presents a comprehensive review of collateral circulation classification in ischemic stroke. • Compares imaging modalities such as CT, MRI, and CBCT for assessing collateral status. • Summarizes manual and automated collateral scoring systems, highlighting their clinical relevance. • Reviews machine learning and deep learning approaches, including CNN, ResNet, VGG, SVM, Random Forest, and k-NN models. • Highlights the effectiveness of multimodal fusion and hybrid architectures in improving classification accuracy. • Discusses current challenges and outlines future directions for clinically applicable AI-based systems.