Skip to main content
2024 AANS Annual Scientific Meeting Main banner
Back

Cerebrovascular Rapid Fire Abstract

Cerebrovascular

Correlation of Aneurysm Dome Diameter to Automated Lesion Detection with Viz.ai

Sunday, May 5, 2024
3:51 PM – 3:53 PM CT
Location:  S405a
Disclosure(s):
Lauren R. Miller, BA: No financial relationships to disclose
David Dornbos, MD, III: Imperative Care: Consultant (Ongoing); Siemens: Grant/Research Support (Ongoing)
Justin F. Fraser, MD: Cerelux: Ownership Interest (Ongoing); Cerenovus: Consultant (Ongoing); Fawkes Biotechnology: Stock Shareholder (excluding mutual funds) (Ongoing); Imperative Care: Grant/Research Support (Ongoing); LETSGETPROOF: Stock Shareholder (excluding mutual funds) (Ongoing); Medtronic: Consultant (Ongoing); NIH: Grant/Research Support (Ongoing); Penumbra: Consultant (Ongoing); Stream Biomedical: Consultant (Ongoing); University of Kentucky: Research Grant (Ongoing)
Introduction: Viz Aneurysm is an artificial intelligence engine created to detect aneurysms from CT-Angiography scans. Currently, Viz.ai recommends that the algorithm be used to detect aneurysms measuring 4 mm or greater. The purpose of this study was to evaluate this algorithm by comparing the maximal dome diameters of aneurysms that were flagged and unflagged by Viz.ai.

Methods: The algorithm analyzed CTA scans at our Comprehensive Stroke Center from October 2022 to June 2023. Information on each aneurysm was collected from Epic, and the radiologist’s determination of maximal dome diameter was used. All aneurysms were divided into a ‘flagged’ or ‘unflagged’ group, according to the algorithm’s detection of an aneurysm. The results were analyzed using GraphPad Prism 9, and the diameters within both groups were compared using the two-tailed Welch’s t test (p < .001).

Results: Of 963 scans analyzed, a total of 70 aneurysms were found, and 45 were detected by the algorithm. Most scans containing aneurysms were performed on females, and patients ranged in age from 34 to 94 years. The minimum threshold for detection was 1 mm, and a total of 16 flagged aneurysms measured less than 4 mm. We found a significant difference (p < .001) between the diameters of the aneurysms detected by the algorithm and those that were not detected. Furthermore, the ‘flagged’ group showed a much larger diameter range (1-22 mm) compared to the ‘unflagged’ group (0.8-10 mm). Overall, the algorithm was significantly more accurate in detecting larger aneurysms.

Conclusion : The present study has found that Viz Aneurysm is more likely to detect aneurysms as their diameters increase. This is an important limitation of this technology because approximately half of the aneurysms in our study were smaller than 4 mm. The clinical relevance of that limitation may be debated, as small unruptured aneurysms are often not treated. Therefore, while this technology provides real-time data feedback to ‘notify’ providers regarding needed attention to an aneurysm, there remain opportunities for further refinement.