Single site performance of AI software for stroke detection and triage

Background: Recently developed software utilizing artificial intelligence for fast detection and triage of stroke cases has the potential to accelerate stroke care and improve patient outcomes. We performed this analysis to evaluate the performance and time to notification of one such software RAPID LVO. Methods: We created a consecutive database of 151 patients for whom scans were processed by the software over a period of eight months. The outputs and running times of the software were collected, alongside patient information based on full patient imaging and evaluation. Results: RAPID LVO achieved a low sensitivity of 63.6% and moderate specificity of 85.8%, with average time to notification of 32.5 minutes, limiting the clinical utility of the alerts. Conclusions: RAPID LVO has low sensitivity, moderate specificity and high time-to-notification performance. Our study data demonstrated in particular low overall sensitivity (63%) for distal occlusions (M2-3). The disparity between the observed performance and the performance reported in RAPID LVO FDA clearance demonstrates the importance of independent, multi- center evaluation. The gap between the performance in this study compared to published records of RAPID AI may be due to differences in imaging hardware, software implementation, connectivity or clinical definitions.