Introduction/Purpose Acute ischemic stroke (AIS) due to large vessel occlusion (LVO) demands rapid and precise diagnostic imaging to guide timely interventions such as endovascular therapy (EVT). Artificial intelligence (AI) has revolutionized stroke imaging by automating detection, segmentation, and outcome prediction, enhancing clinical decision‐making. This systematic review maps the scope of AI‐driven imaging modalities for stroke diagnosis, identifying current technologies, their applications, and critical gaps to guide future development and clinical integration. Materials and Methods A comprehensive analysis of 49 peer‐reviewed studies from 2019 to 2025 was conducted, sourced from a curated dataset focusing on AI applications in AIS imaging. Studies included retrospective and prospective cohorts evaluating AI tools for LVO detection, infarct segmentation, and outcome prediction using modalities such as computed tomography angiography (CTA), CT perfusion (CTP), non‐contrast CT (NCCT), and novel optical blood flow monitoring. Performance metrics (sensitivity, specificity, area under the receiver operating characteristic curve AUROC, Dice similarity coefficient DSC) and clinical workflow impacts (e.g., time‐to‐notification) were extracted. Technologies were categorized by imaging modality, AI approach (e.g., convolutional neural networks CNNs, deep learning), and clinical application. Gaps were identified through qualitative synthesis of study limitations and reported challenges. Results AI‐driven tools demonstrated robust performance across multiple modalities. CTA‐based CNNs (e.g., Viz LVO, CINA LVO) achieved sensitivities of 74.6‐100% and specificities of 71‐97% for proximal LVO detection (ICA, M1), with AUROCs of 0.74‐0.95, though M2 occlusion detection was less reliable (0‐58% sensitivity). NCCT‐based algorithms detected hyperacute stroke signs with sensitivities comparable to neuroradiologists (77‐93%). CTP and multiphase CTA models predicted infarct volumes with correlations up to r=0.92 and DSCs of 0.35‐0.85, rivaling traditional methods. Novel modalities, like the Openwater optical blood flow monitor, outperformed prehospital scales (AUROC 0.82 vs. 0.65‐0.70). AI reduced diagnostic delays by 7‐26 minutes and improved EVT triage. Key gaps included limited sensitivity for distal occlusions, poor model interpretability, variability in low‐prevalence settings (e.g., 67% false discovery rate at 4.1% prevalence), and insufficient large‐scale public datasets. Conclusion AI‐driven imaging modalities, including CTA, CTP, NCCT, and emerging optical technologies, significantly enhance stroke diagnosis by improving LVO detection, infarct prediction, and workflow efficiency. However, gaps in distal occlusion detection, model transparency, and generalizability across diverse populations necessitate further research. Future efforts should prioritize large public datasets, interpretable algorithms, and integration of novel modalities to address these gaps and optimize AI's role in stroke care.
Elsayed et al. (Sat,) studied this question.