Catch More or Call Fewer? A Systematic Review and Meta-analysis comparing Viz.ai vs Rapid AI
Hafiz Talha Javed1, Sama Almasri1, Divya Nayar1, MOHAMMAD ALKHALDI1, Muhammad Hassan Akhtar2, Ali Hamdan2, Zainab Jamshaid2, Krishna Nalleballe1, Sanjeeva Reddy Onteddu1
1Department of Neurology, University of Arkansas for Medical Sciences, 2Department of Neurology, King Edward Medical University
Objective:
We conducted a systematic review and meta-analysis comparing the diagnostic performance of Rapid AI and Viz.ai for LVO detection.
Background:
Management of acute ischemic stroke due to large-vessel occlusion (LVO) is highly time-sensitive. Artificial-intelligence tools such as Rapid AI and Viz.ai are increasingly used to facilitate rapid detection and prioritization of LVO on brain CT angiography (CTA).
Design/Methods:
A PRISMA-guided search of PubMed and the Cochrane Library from inception to August 28, 2025, was performed. Analyses were conducted in R (meta, metafor) using random-effects models. The primary outcomes were pooled sensitivity and specificity for LVO detection. Secondary outcomes included LVO detection rate (n/N), positive predictive value (PPV), negative predictive value (NPV), and false-positive rate (FPR).
Results:
Eighteen studies included 14,387patients, met inclusion criteria (13 Rapid AI; 5 Viz.ai). Rapid AI showed higher sensitivity (0.85; 95% CI, 0.80–0.90) and a higher detection rate (0.88; 0.75–0.95) versus Viz.ai (sensitivity 0.75; 0.66–0.83; detection rate 0.74; 0.66–0.81). Viz.ai demonstrated higher specificity (0.95; 0.92–0.98) than Rapid AI (0.87; 0.83–0.91). PPV was 0.57 (0.48–0.65) for Rapid AI and 0.65 (0.52–0.77) for Viz.ai; NPV remained high for both (Rapid AI 0.98 [0.97–0.98]; Viz.ai 0.97 [0.96–0.99]). FPR was lower for Viz.ai (0.05; 0.02–0.08) than Rapid AI (0.10; 0.05–0.19). All pooled estimates were statistically significant (p<0.001).
Conclusions:
Rapid AI prioritized case capture (higher sensitivity/detection), whereas Viz.ai prioritized precision (higher specificity/lower FPR). Choice of tool may be tailored to institutional priorities. Centers prioritizing maximal LVO capture may favor Rapid AI, whereas those aiming to limit false alerts and downstream evaluations may prefer Viz.ai.
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.