Segmentation and Modeling for Accurate Reconstruction of CT Angiography of Intracranial Large Vessel Occlusion with Artificial Intelligence: a Stepped-wedge, Cluster-randomized Controlled Trial

Acute ischemic stroke (AIS) caused by intracranial large vessel occlusion (LVO) in the anterior circulation significantly contributes to stroke-related disability and mortality. Recent randomized controlled trials have demonstrated substantial benefits of endovascular thrombectomy (EVT) when patients are appropriately triaged beforehand. However, accurately orienting the 'missed segment' during EVT remains challenging. Guide-wires often fail to navigate through the occlusion or are mistakenly directed into the small tranches or even cause vessel rupture. To address this clinical need, the investigators developed an artificial intelligence (AI) algorithm to automate the reconstruction of CT angiography (CTA), focusing on the occluded LVO segment. To evaluate the clinical utility of this AI algorithm, the investigators propose a prospective, stepped-wedge cluster-randomized study to determine whether integrating our AI algorithm into AIS care flow can reduce the time for first pass of the thrombus by improving the visualization of the occluded segment on CTA. Physicians will assess patient eligibility for thrombectomy, and all selected patients will receive standard care according to current guidelines. This approach is expected to enhance patient treatment outcomes for endovascular thrombectomy by leveraging readily available data.