Accuracy of RapidAI’s Aneurysm Platform in Detection of Intracranial Aneurysms: Retrospective Analysis of Consecutive Series of 11,694 Computed Tomography Angiograms (CTA).

Introduction: Unruptured intracranial aneurysms (UIAs) have an estimated prevalence of 3-5%. Early detection of UIAs, and identification of high-risk patients is vital part of planning for preventive intervention. Rapid Aneurysm is an Artificial Intelligence platform which detects and measures cerebral aneurysms with 3D reconstructions. In this study we aimed to compared Rapid Aneurysm’s diagnostic results to radiology reports in a large sample of CTAs.

Methods: A cohort of 11,694 consecutive CTA studies completed between September 2020 and June of 2024 was analysed retrospectively. 61 studies were excluded due to missing reports, or images, insufficient scan length, duplication, or artifacts. All scans were independently assessed by the Rapid Aneurysm software. Radiologists were blinded to Rapid’s results. All suspected aneurysms, determined either by Rapid or by the radiologist, were subsequently reviewed by two senior radiology residents. Aneurysms less than 2mm n=15), and those located outside of the anatomical regions of segmentation (n=14)were considered true negative.

Results: Total of 11,633 were included in analysis. 1206 studies flagged by the software and/or the radiologist as having a potential aneurysm underwent clinical review. One or more confirmed aneurysms was present in 524 (4.5%) studies. Rapid Aneurysm correctly identified 460/524 aneurysms, sensitivity 87.8% (CI: 0.85-0.90) and 10,863/11,109 aneurysm-negative studies, specificity 97.8%, (CI: 0.97-0.98), positive predictive value 65.1% (CI:0.61%-0.69%), and negative predictive value 99.4% (CI: 0.99.2-0.99.5). Radiologists detected 372 aneurysms. Rapid identified 150 aneurysms not mentioned in the original radiology report (median max dimension: 3.9mm [2.26-13.8mm]) while 63 aneurysms missed by the software were detected by radiologists (median max dimension: 3mm [2-5mm]).

Conclusion : This study demonstrates that AI not only achieves high accuracy in detecting IAs but also identified a substantial number of aneurysms that were missed by radiologists. These findings underscore AI’s potential role in enhancement of diagnostic accuracy, and proper and management of IAs.