Aidoc has announced the presentation of two new scientific abstracts comparing artificial intelligence (AI)-powered stroke solutions at this year’s International Stroke Conference (ISC; 4–6 February, New Orleans, USA). The company states that the data—presented by independent academic institutions—highlight the real-world reliability of Aidoc’s neuro AI algorithms and demonstrate a “significant performance advantage” over other market solutions, particularly in the triage of medium-vessel occlusions (MeVOs).
A comparative study conducted by the University of Texas Medical Branch (Galveston, USA) evaluated the performance of Aidoc against what the company describes as a “conventional AI vendor” in a clinical setting, contrasting the triage accuracy of both systems across 1,557 consecutive computed tomography angiography (CTA) exams.
The results revealed higher large vessel occlusion (LVO) sensitivity for head-to-head LVOs in the internal carotid artery (ICA) or proximal M1 segment of the middle cerebral artery (MCA) with Aidoc’s technology, as per a rate of 92.6% compared to 70.4% with the conventional solution.
Aidoc also highlights a “distinct performance gap” regarding the flagging of MeVOs, with the company’s solution—which has gained US Food and Drug Administration (FDA) clearance for analysis in a greater number of total segments—achieving a 92.3% sensitivity rate compared to 17.3% with the conventional AI product.
Additionally, in terms of overall reliability, Aidoc’s technology achieved a negative predictive value (NPV) of 99.3%, which the company claims should give clinicians an increased level of confidence in the notifications provided by the solution.
“These findings illuminate a crucial distinction in the market: simply having an algorithm isn’t the same as delivering clinical-grade performance,” said Elad Walach, co-founder and chief executive officer (CEO) of Aidoc. “When we see a 75% gap in sensitivity for identifying MeVOs, it underscores that high-fidelity AI is a necessity for ensuring that the complete spectrum of complex, time-sensitive cases are flagged for immediate review.”
In a separate study presented via the Saint Louis University School of Medicine (St Louis, USA), researchers analysed 1,000 CTA exams to evaluate the effectiveness of Aidoc’s vessel occlusion algorithm in a “messy” real-world environment. The study demonstrated that high fidelity performance supports AI integration into workflows for evaluating both anterior- and posterior-circulation occlusions, according to Aidoc.
The study produced high specificity, as Aidoc’s solution maintained a rate of 97.1%, in addition to broad vessel coverage—with the algorithm successfully flagging occlusions in both the anterior circulation (sensitivity, 87.6%) and the more challenging posterior circulation (sensitivity, 79.2%).
