RapidAI may be more accurate than VizAI in detecting large vessel occlusion
Key takeaways:
- The RapidAI software successfully processed and identified more eligible large vessel occlusion cases vs. VizAI.
- It also correctly identified more LVO-negative cases and detected more LVOs missed by VizAI.
RapidAI automated imaging appeared to be more accurate than VizAI in detecting cases of large vessel occlusion, as well as in identifying those missed by its counterpart among individuals with acute stroke, results of a comparative analysis show.
“When it comes to stroke care, ‘time is brain’ and the accuracy of clinical AI tools is critically important,” Harmeet Sachdev, MD, lead study author and neurologist at Good Samaritan Hospital in San Jose, California, told Healio of the study presented at the International Stroke Conference. “As clinical AI becomes increasingly utilized in patient care, we sought to understand what difference, if any, could be observed between two leading [large vessel occlusion] detection software tools.”
Sachdev and colleagues compared the performance of RapidAI and VizAi in the simultaneous detection of large vessel occlusion (LVO) among adults with stroke who were treated at a comprehensive stroke center.
Their retrospective analysis culled data from almost 1,600 instances of stroke found on computed tomography angiography (CTA) between June 2022 and June 2024. An LVO diagnosis was confirmed by radiology reports and expert review of both CTA and CT perfusion imaging.
From the 1,525 cases suitable for analysis (average age, 67.8 years; 53.5% women), 1,376 were LVO negative and the remaining were LVO positive. Diagnostic accuracy was then compared for both AI-based technologies in processing both LVO positive and negative cases.
According to the results, RapidAI software successfully processed 1,523 (99%) eligible stroke cases, while VizAI successfully processed 1,432 (90%).
RapidAI also outperformed VizAI in correct identification of LVOs (98% vs. 74%) as well as LVO-negative cases (94% vs. 91%).
Data further showed that RapidAI was able to detect 30 LVOs missed by VizAI, while VizAI was only able to successfully detect three LVOs missed by RapidAI.
“This calls attention to the importance of understanding the sensitivity and specificity of clinical software to avoid the risk of delays in diagnosis and treatment of patients with acute ischemic stroke,” Sachdev told Healio.