Article Text

Download PDFPDF

E-117 Expanding horizons: Viz.AI’s versatile role in neurological care
  1. E Paulin1,
  2. D Williams-Stankewicz1,
  3. J Frank2,
  4. N Millson2,
  5. M Campbell3,
  6. L Wise3,
  7. D Lukins4,
  8. M Al-Kawaz5,
  9. D Dornbos2,
  10. S Pahwa4,
  11. J Fraser2
  1. 1College of Medicine, University of Kentucky, Lexington, KY
  2. 2Neurosurgery, University of Kentucky, Lexington, KY
  3. 3Neurology, University of Kentucky, Lexington, KY
  4. 4Radiology, University of Kentucky, Lexington, KY
  5. 5Interventional Neurology, University of Kentucky, Lexington, KY

Abstract

Objective Viz.AI is an artificial intelligence program designed to screen patients with suspected large vessel occlusions (LVO) and other cerebrovascular emergencies. This study aims to compare the sensitivity and specificity of Viz.AI with the diagnostic skills of radiologists at a Comprehensive Stroke Center. Additionally, an analysis of factors contributing to Viz.AI’s predictions was conducted.

Methods Scans were processed through Viz.AI’s LVO detection platform from October 23, 2022, to September 10, 2023. Radiology reads of all scans were reviewed using the electronic medical record system, EPIC. Data were analyzed using Prism 9.0.

Results Among the 1,409 scans recorded, 238 were flagged as LVO by Viz.AI, with 143 confirmed as LVO by radiologists, resulting in 87 cases receiving endovascular treatment. Of the 95 LVO cases flagged by Viz.AI not confirmed by review, 43 (45%) required a Neurosurgery consult for other cerebrovascular pathology (moyamoya, hematoma, hemorrhage, encephalopathy, prior stroke, stenosis, occlusion of a medium-sized vessel, ICAD, and cysts). Statistical analysis revealed a sensitivity of 71.86% (95% CI 65.06–77.99), specificity of 92.12% (95% CI 90.46–93.58), positive predictive value of 60.08% (95% CI 54.92–65.04), and a negative predictive value of 95.20% (95% CI 94.08–96.12).

Conclusions The study underscores the wide-ranging impact of Viz.AI’s AI technology in revolutionizing critical care. Its LVO detection platform, functioning as a versatile tool, demonstrates sensitivity beyond large vessel occlusions, extending to various cerebrovascular emergencies and other pathological conditions. Recognizing this versatility and addressing limitations can enable healthcare providers to fully utilize Viz.AI as a vital tool in diagnosing and managing neurological disorders.

Disclosures E. Paulin: None. D. Williams-Stankewicz: None. J. Frank: None. N. Millson: None. M. Campbell: None. L. Wise: None. D. Lukins: None. M. Al-Kawaz: None. D. Dornbos: None. S. Pahwa: None. J. Fraser: None.

Statistics from Altmetric.com

Request Permissions

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.