Document resource
Introduction/Purpose Medium Vessel Occlusions (MeVOs) account for nearly 40% of acute ischemic strokes and are associated with poor patient outcomes. However, identifying MeVOs is challenging due to their small size and highly variable vascular anatomy. This study evaluates the utility of the RapidAI platform to automate the detection of these occlusions on Computed Tomography Angiography (CTA) and Computed Tomography Perfusion (CTP) imaging.Materials and Methods 133 acute stroke patients with radiology adjudicated confirmation of a MeVO and who had undergone both CTA and CTP imaging were identified from retrospective databases at MUSC, Stanford, and Royal Melbourne. MeVOs were defined as occlusions in the MCA (M2-M3), ACA (A2-A3), or PCA (P1-P3) segments, with each location confirmed by an expert neuroradiologist. Nineteen patients were excluded due to technically inadequate imaging, such as excessive patient motion, metal artifacts, or poor contrast bolus. The remaining scans were post-processed using four automated software modules: Rapid LVO which alerts teams to large vessel occlusions (ICA/M1), Rapid CTA Vessel Density which assesses hemispheric blood vessel density relative to the contralateral side, Rapid LMVO which alerts teams to large and medium vessel occlusions, and Rapid CTP which identifies likely irreversibly injured tissue and potentially salvageable brain tissue (iSchemaView, Inc., Menlo Park, CA).Results 114 cases were included in the final analysis. The combination of Rapid LVO, CTA Vessel Density, and LMVO detected 78.1% (89/114) of MeVOs, a 53.4% improvement (p<0.0001) over using Rapid LVO+CTA alone (50.9%; 58/114)(table 1). The addition of Rapid CTP further increased the overall detection rate to 99.1% (113/114), marking a 26.9% improvement over the Rapid LVO+CTA+LMVO baseline (p=0.1207). The single false negative was a distal M3 occlusion that fell below the threshold for detection. The most significant improvement occurred in the PCA (P1-P3) territory, where detection rose from 23.9% (11/46) on CTA to 100% (46/46) after adding Rapid LMVO and CTP (p<0.0001). High detection rates were also achieved in the MCA (M2) segment at 100% (54/54) and the ACA (A2-A3) segments at 100% (8/8).Conclusion Adding Rapid LMVO and Rapid CTP significantly improves MeVO detection across all brain regions, especially in the posterior circulation. For hospitals that do not perform perfusion imaging, Rapid LMVO serves as a critical tool for the timely identification and triage of these patients. Overall, a multi-modal automated workflow ensures that more distal occlusions are caught early, allowing for faster treatment and better patient care.Disclosures M. Sowlat: None. A. Zamarud: None. B. Campbell: None.