An Artificial Intelligence (AI)-Based Recruitment Tool for Patient Identification and Enrollment in the MEMBRANE trial

Introduction: The MEMBRANE trial (NCT 04816591) assesses middle meningeal artery (MMA) embolization's safety and efficacy for chronic subdural hematoma (cSDH). This report focuses on a comprehensive stroke center's experience, utilizing the Viz RECRUIT SDH AI tool to identify and enroll cSDH patients based on non-contrast computed tomography (NCCT) images.

Methods: The MEMBRANE trial began on August 26, 2022, concurrently with the activation of the AI-based recruitment tool, Viz RECRUIT SDH (Viz.ai Inc., San Francisco, CA), on August 24, 2022. This tool detects cSDH on NCCT, alerts the research team, and facilitates image viewing and messaging. The tool's performance was evaluated from August 25, 2022, to May 25, 2023, encompassing 6076 head NCCTs. Primary endpoint: percentage of enrolled patients identified by the tool. Secondary endpoints: percentage of viewed AI alerts, cSDH detections, and radiographic characteristics (volume, thickness, midline shift) of suspected cSDH cases.

Results: Out of 6076 NCCT images, 5844 (96.2%) were successfully processed. The median [IQR] age was 65.0 [44.5 – 77.0], with 39% female. All 6 enrolled patients were identified by the tool. 451 (7.7%) alerts were sent; 372 (82.5%) were viewed. Within an hour, 39.5% alerts were reviewed, and 83.3% within a day. Median [IQR] volume, thickness, midline shift were 24.0 [55.5] mL, 10.0 [7.0] mm, 1.0 [3.0] mm, respectively.

Conclusion : An effective AI-based recruitment tool, offering automatic cSDH detection, image viewing, and team messaging, enhances the identification and enrollment of cSDH patients in pivotal clinical trials like MMA embolization in large comprehensive stroke centers.