Introduction/Purpose Endovascular thrombectomy has been shown to decrease disability, improve functional outcomes, and reduce mortality over standard medical management for anterior Large Vessel Occlusion (LVO) within a time-sensitive window. The window for thrombectomy is limited, and LVOs can produce large infarcts if timely intervention is not performed to salvage ischemic brain tissue. Given the time-sensitive nature of these interventions, computer automation may aid in the early detection of these LVOs, but false values may lead to alarm fatigue and ignoring these systems altogether. We compared the accuracy of two commercially available products, Viz LVO and Rapid LVO, to detect these LVOs.

Materials and Methods Data was taken retrospectively from two HCA Houston facilities from January 2022-January 2023 from multimodal computed tomography with thin-slice computed tomography angiography (CTA) for suspected LVO. Two automated LVO detection packages, Rapid LVO and Viz LVO, were run on these images, and a Diagnostic Neuroradiologist or Neurointerventional Radiologist interpreted the scans. We calculated the sensitivity, specificity, positive predictive value, and negative predictive value and performed a McNemar test to look for a difference between the algorithm’s classifications. On each of these patients, we also collected demographic data, comorbidities, ejection fraction (EF), and intracranial atherosclerosis as defined by radiologist interpretation and verification by Modified Woodcock Score >1. A multinomial logistic regression was performed to determine if any of these variables predicted the incorrect classification of a CTA.

Results A total of 360 participants were included with a mean age of 65 years old with a standard deviation of +/-16.5 with 159 males and a total of 47 large vessel occlusions confirmed by diagnostic or neurointerventional radiologists. Rapid LVO had a specificity of 0.85 and a sensitivity of 0.87, with a positive predictive value (PPV) of 0.46 and a negative predictive value (NPV) of 0.97. Viz LVO had a specificity of 0.96 and a sensitivity of 0.87, with a PPV of 0.75 and a NPV of 0.98. We formed a contingency table of correct and incorrect classification and performed a McNemar test showing a statistically significant difference between classifications by the two algorithms (p=0.00000031). We took the data relating to demographics, comorbidities, atherosclerosis on imaging, and EF and performed a multinomial logistic regression on the incorrect predictions of the algorithms. On Viz LVO, low EF (p=0.00125) and Modified Woodcock Score >1 (p=0.000198) were significant predictors of incorrect classification. When using Rapid LVO, EF (p=0.0286) and Modified Woodcock Score >1 (p=0.000000975) were also significant predictors of incorrect classification.

Conclusion Viz LVO and Rapid LVO had similar NPV, but Rapid LVO produced a significantly larger number of false positive values. False positive values can be a source of alarm desensitization, leading to missed alarms or delayed responses. EF and intracranial atherosclerosis were significant predictors of incorrect predictions in both software packages. We hope this data will be used to improve future algorithms.

Disclosures A. Delora: None. C. Hadjialiakbari: None. E. Percenti: None. A. Hassan: 1; C; LVO SYNCHRONIZE-Viz.ai. 2; C; Viz.ai. Y. Alderazi: None. J. Torres: None. M. Ezzeldin: None.