Construction and Evaluation of an Intelligent Decision System for Reperfusion Therapy in Acute Ischemic Stroke

The LingBao System (Stroke Reperfusion Intelligent Decision System, SRIDS) is an AI-based clinical decision support platform designed to assist physicians in the evaluation and treatment decision-making process for acute ischemic stroke (AIS) patients eligible for reperfusion therapy. Built upon the 2024 Chinese Guidelines for Reperfusion Therapy in Acute Ischemic Stroke, major international randomized controlled trial (RCT) evidence, and large-scale real-world datasets, LingBao integrates clinical information, imaging parameters, and guideline-based criteria to provide transparent, evidence-graded recommendations for intravenous thrombolysis and/or endovascular therapy.

This multicenter, cluster-randomized controlled trial aims to assess the real-world effectiveness and safety of LingBao in optimizing reperfusion decision-making and improving patient outcomes. Twenty certified stroke centers in China will be randomized 1:1 to either LingBao-assisted care or standard care. Consecutive adult patients (≥18 years) presenting within 24 hours of last-known-well and evaluated for reperfusion therapy will be prospectively enrolled.

The initial target sample size is approximately 3,000 participants enrolled from about 20 certified stroke centers randomized in a 1:1 ratio. This target accounts for the cluster-randomized design and the anticipated intracluster correlation. To preserve adequate statistical power in the presence of uncertainty regarding the intracluster correlation, unequal cluster sizes, follow-up completion, and the overall distribution of the 90-day modified Rankin Scale, the protocol includes a pre-specified blinded sample size re-estimation. The blinded re-estimation will be conducted once approximately 40% of participants have completed the 90-day primary outcome assessment. An independent statistician will perform this analysis using pooled data without access to treatment allocation. The re-estimation will not include any between-group comparison or treatment-effect estimate. Parameters considered will include the actual mean cluster size, variability in cluster size, intracluster correlation, 90-day follow-up completion, missingness of the primary outcome, and the overall mRS distribution. If the updated design effect or other blinded assumptions indicate that the current sample size may provide insufficient statistical power, enrollment may be increased. Any adjustment will be limited to an increase in sample size and will be implemented by extending recruitment at existing centers or, if necessary, adding additional certified stroke centers. The procedure is designed to maintain the integrity of randomization and control the type I error rate.

At intervention sites, clinicians may use LingBao during standard clinical workflows. The system automatically calculates the estimated onset-to-treatment window, screens contraindications, summarizes imaging-based eligibility criteria, and displays guideline-concordant recommendations along with their class and level of evidence. LingBao does not replace physician judgment; all clinical decisions remain entirely at the discretion of the treating team.

Data on patient demographics, baseline characteristics, workflow times, treatments, and outcomes will be collected through standardized electronic case report forms (eCRFs). The primary outcome is functional status at 90 days, measured by the modified Rankin Scale (mRS) and analyzed using an ordinal shift model. Key secondary outcomes include door-to-needle time (DNT) for intravenous thrombolysis, door-to-puncture time (DPT) for endovascular therapy, rates of reperfusion treatment, early neurological improvement, symptomatic intracranial hemorrhage, and mortality.

By systematically integrating AI-driven clinical reasoning with evidence-based medicine, the LingBao study aims to establish an intelligent, reproducible, and guideline-concordant framework for acute stroke management. The results are expected to inform large-scale implementation of AI-supported decision systems to enhance the quality, consistency, and efficiency of stroke reperfusion therapy in real-world practice.