AI software detection of large vessel occlusion stroke on CT angiography: a real-world prospective diagnostic test accuracy study

Abstract

BackgroundArtificial intelligence (AI) software is increasingly applied in stroke diagnostics. However, the actual performance of AI tools for identifying large vessel occlusion (LVO) stroke in real time in a real-world setting has not been fully studied.ObjectiveTo determine the accuracy of AI software in a real-world, three-tiered multihospital stroke network.MethodsAll consecutive head and neck CT angiography (CTA) scans performed during stroke codes and run through an AI software engine (Viz LVO) between May 2019 and October 2020 were prospectively collected. CTA readings by radiologists served as the clinical reference standard test and Viz LVO output served as the index test. Accuracy metrics were calculated.ResultsOf a total of 1822 CTAs performed, 190 occlusions were identified; 142 of which were internal carotid artery terminus (ICA-T), middle cerebral artery M1, or M2 locations. Accuracy metrics were analyzed for two different groups: ICA-T and M1 ±M2. For the ICA-T/M1 versus the ICA-T/M1/M2 group, sensitivity was 93.8% vs 74.6%, specificity was 91.1% vs 91.1%, negative predictive value was 99.7% vs 97.6%, accuracy was 91.2% vs 89.8%, and area under the curve was 0.95 vs 0.86, respectively. Detection rates for ICA-T, M1, and M2 occlusions were 100%, 93%, and 49%, respectively. As expected, the algorithm offered better detection rates for proximal occlusions than for mid/distal M2 occlusions (58% vs 28%, p=0.03).ConclusionsThese accuracy metrics support Viz LVO as a useful adjunct tool in stroke diagnostics. Fast and accurate diagnosis with high negative predictive value mitigates missing potentially salvageable patients.

Artificial Intelligence-Assisted Software Significantly Decreases All Workflow Metrics for Large Vessel Occlusion Transfer Patients, within a Large Spoke and Hub System

“…AI-assisted software is one tool to streamline acute stroke transfer processes. Its potential application is broad, from assessment of ASPECTS score to identification of salvageable tissue with perfusion imaging and LVO detection [7, 15]. Some AI software, including Viz LVO, includes tools for rapid communication as well.…”

“…Some AI software, including Viz LVO, includes tools for rapid communication as well. We have previously demonstrated the high sensitivity and specificity of AI software in detecting LVO [7].…”

“…Viz LVO is trained to identify LVOs in the supraclinoid internal carotid artery (ophthalmic, choroidal, and communicating segments) and the M1 (horizontal part) of the MCA. However, it does not assess the extracranial circulation, the posterior circulation, or the infraclinoid internal carotid artery [7]. In instances where a partial or complete occlusion is suspected, or when a vessel's caliber is less than the reference threshold, an LVO is suspected, and an alert is automatically sent to the stroke team [8].…”

“…Viz LVO was authorized by the Food and Drug Administration in 2018 to assist with LVO detection in computed tomography angiography (CTA). The high sensitivity and negative predictive value of this AI-driven software might help decrease the rate of underdiagnosed patients who would benefit from EVT [7]. The application includes features that enable early team notification and communication between the stroke team members, and early diagnosis and team notification may decrease workflow time [7, 8].…”

Recent developments in pre-hospital and in-hospital triage for endovascular stroke treatment

“…Advanced visualization technologies and automation of image processing and analysis could be of particular value for stroke triage in smaller hospitals (eg, PSCs with limited neuroimaging experience), or hospitals in which trainees are the primary image readers off-hours. Examples for such (semi-)automated image analysis modules include automated hemorrhage detection,60 61 ASPECTS (Alberta Stroke Program Early CT Score) scoring,62–65 LVO detection,66 67 and generation of perfusion maps from multiphase CTA 68–70. These technologies could transform the acute stroke imaging workflow by (1) improving diagnostic accuracy (reduced risk of misdiagnosis), (2) timely ‘rescue’ of patients who have been missed in the earlier parts of the pre-hospital workflow (eg, patients with atypical or mild symptoms for whom no code stroke was initiated), and (3) linking automated detection of stroke imaging findings with instant notification of physicians and stroke teams, which could enable rapid concerted responses.…”