Abstract WMP67: Evaluation Of The Implementation Of An Ai Tool For Large Vessel Occlusion: Impact On Radiologists’ Workflow And Patient Outcomes

• DOI: 10.1161/str.53.suppl_1.wmp67
• Publication Date: 2022-02-03T10:01:59Z
• Authors: Masaki Nagamine; Anna Zolyan; Sebastian Elstrott; Joel Mclouth; Viet Vu; Wengui Yu; Jennifer Soun
• Keywords: Triage; Stroke; Turnaround time; Acute stroke

Timely and accurate diagnosis of a large vessel occlusion (LVO) is critical for prompt initiation thrombectomy treatment to reduce morbidity mortality in acute ischemic stroke patients. Recently, artificial intelligence (AI) tools aid triage have shown promise their accuracy speed. However, little known about the impact clinical integration AI on efficiency metrics patient outcomes. This study aims evaluate an AI-based LVO detection tool three main areas: (1) performance metrics, (2) radiologist workflow, (3) single center retrospective analysis includes patients with suspected who underwent CT angiograms (CTA) month period before after implementation tool. CTA 224 pre-implementation 144 post-implementation were evaluated. The correctly identified 93/107 cases, 87%, sensitivity 91%, specificity 86%. PPV was 0.64 NPV 0.97. mean radiology report turnaround time (TAT) using significantly reduced at 19.9 minutes (SD 32.4) compared 33.7 34.7) without (p=0.001). There significant improvement resident trainee TAT tool, 17.9 32.6) 38.4 33.6) (p=0.0002). Baseline risk factors similar between pre groups. no differences metric medians door-to-image, door-to-needle, door-to-puncture, door-to-revascularization times there statistically NIHSS arrival, post NIHSS, discharge mRS, or mortality. findings outcomes may be attributable small sample size. validates demonstrates associated reduction TAT, allowing faster communication ordering clinicians. Further research needed these benchmarks establish role