Abstract Details

Title Real World Experience with Viz.ai Automated Hemorrhage Detection at a Comprehensive Stroke Center

Topic Cerebrovascular Disease and Interventional Neurology

Presentation(s) P1 - Poster Session 1 (8:00 AM-9:00 AM)

Poster/Presentation
Number 5-023

Objective

To determine the accuracy of Viz.AI Automated Hemorrhage Detection at our large academic comprehensive stroke center.

Background The use of artificial intelligence (AI) to assist in acute ischemic stroke imaging has recently gained widespread acceptance. Viz.AI added an automated detection of suspected hemorrhage algorithm to its platform. Real-world performance of Viz.AI hemorrhage detection is limited. We report our experience with Viz.AI hemorrhage detection at our academic Comprehensive Stroke Center (CSC).

Design/Methods We performed a retrospective review of stroke alerts with non-contrast CT head (CTH) performed from May 12, 2021 to September 19, 2021. Data was collected on sex, age, Viz.AI hemorrhage alert, alert times, and CTH findings. Radiologist determination of hemorrhage on CTH was considered gold standard. Intracranial hemorrhage diagnosis included intracerebral (ICH), subarachnoid (SAH), and subdural hemorrhages (SDH). Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated.

Results Among 801 consecutive suspected stroke patients with non-contrast CTH analyzed by Viz.AI, the mean age was 64.1 years (range 17.8-97.4), and 401 (50%) were female. Viz.AI hemorrhage autodetection alerted in 113 (14.1%) CTH scans. Radiology review reported 118 patients with intracranial hemorrhage. There were 670 true negatives, 13 false positives, 18 false negatives, and 100 true positives. The 18 FNs (Viz AI hemorrhage detection did not alert in the presence of hemorrhage) included 4 SDH, 7 SAH (6 convexity and 1 interpeduncular), and 7 ICH. Sensitivity of ICH autodetection was 84.7%, specificity 98.1%, PPV 88.5% and NPV 97.4%. The mean time from CT completion to Viz.AI hemorrhage alert notification was 2.1 minutes (range 1-7 minutes).

Conclusions In our series of patients evaluated for stroke code at our CSC, Viz.AI automated hemorrhage detection performed well with a sensitivity of 84.7%, specificity of 98.1%, and mean alert time of 2 minutes. Further study is warranted to improve processes in acute hemorrhagic stroke care.

Authors/Disclosures
Nick Mannix, MD (The Ohio State University Wexner Medical Center) PRESENTER	Dr. Mannix has nothing to disclose.
Narsis Aminian, MD	Dr. Aminian has nothing to disclose.
Shahid Nimjee, MD, PhD (The Ohio State University Wexner Medical Center)	Dr. Nimjee has received personal compensation in the range of $500-$4,999 for serving as a Consultant for Silvercreek. Dr. Nimjee has received personal compensation in the range of $10,000-$49,999 for serving on a Speakers Bureau for Medtronic. Dr. Nimjee has stock in Basking Biosciences. The institution of Dr. Nimjee has received research support from National Institutes of Health.
Mohammad Taimur Shujaat (OSU Wexner Medical Center)	No disclosure on file
Cassandra Forrest (Wexner medical Center)	Cassandra Forrest has nothing to disclose.
Sharon Hammond-Heaton, RN (The Ohio State University Wexner Medical Center)	Ms. Hammond-Heaton has nothing to disclose.
Vivien H. Lee, MD, FAAN (OSU Comprehensive Neurovascular Center)	Dr. Lee has nothing to disclose.

��ɫ��

��ɫ��