Introduction Stroke caused by emergent large vessel occlusion (LVO) is a critical time-sensitive diagnosis requiring prompt identification to identify candidates for endovascular therapy (EVT). As a result, identifying imaging findings on non-contrast computed tomography (NCCT) that are predictive of LVO would aid in the identification of potential EVT candidates. We present and validate gaze deviation as an indicator for detecting LVO using NCCT. In addition, we present an Artificial Intelligence (AI) algorithm for detecting this indicator.
Methods AI algorithms have tremendous potential to aid in this triage process but have so far been limited to brain parenchymal predictors of LVO. We present an AI algorithm to detect gaze deviation from an NCCT scan. The AI algorithm was trained using a set of 200 scans to identify gaze direction. The gaze deviation is calculated by measuring the angle between the gaze direction and the midline of the brain. We used this AI algorithm to identify clinical symptoms of ipsiversive gaze deviation in 116 stroke patients with LVO treated with EVT. This data was gathered at two stroke centers of a neuroscience institute, where it was annotated and validated by experienced neuro-intervention providers.
Results NCCT revealed that 71.1% (59/83) of proximal occlusions had an ipsiversive gaze deviation. In 79% (47/59) of cases, the AI algorithm correctly identified this clinical predictor of proximal LVO. M2 occlusions with less severe clinical symptoms were less likely to show ipsiversive gaze deviation on NCCT, with 42.4% (14/33) of patients showing it. Again, the AI algorithm performed admirably, identifying 85.7% (12/14) of these gaze deviations based solely on NCCT. When tested on normal control scans, the AI algorithm yielded an accuracy of 80.1%. AI algorithm had sensitivity and specificity of 80.8% and 80.1%, respectively.
Conclusions Ipsiversive Gaze deviation on NCCT is a good predictor of LVO due to proximal vessel occlusions in ICA terminus and M1 occlusions. However, it is a poor predictor of LVO due to M2 occlusion. We report an AI algorithm that can identify this clinical sign on NCCT. These findings can aid in the triage of LVO patients and expedite the identification of EVT candidates.
Disclosures U. Upadhyay: None. S. Golla: None. S. Kumar: None. K. Szweda: None. R. Shahripour: None. J. Tarpley: 2; C; Qure.ai – Modest, Medtronic – Modest, Stryker – Modest.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.