Can Artificial Intelligence to Detect Large Vessel Occlusion Improve Patient Care? A Systematic Review and Meta-analysis
Julyana Dantas1, Giovana Ribeiro2, Caroline Dagostin3, Antonio Mutarelli4, Pedro Romeiro5, Giulia Almirón6, Agostinho Pinheiro7
1Universidade Federal do Rio Grande do Norte, 2Unicamp - Universidade Estadual de Campinas, 3Universidade do Extremo Sul Catarinense, 4Universidade Federal de Minas Gerais, 5UNIMA/AFYA - Centro Universitário de Maceió, 6UNIMES - Universidade Metropolitana de Santos, 7Department of Neurology, Massachusetts General Hospital, Brigham and Women’s Hospital, Harvard Medical School
Objective:
We aimed to perform a systematic review and meta-analysis comparing artificial intelligence (AI)-assisted radiological assessment versus standard assessment of large vessel occlusion (LVO) in patients with acute stroke.
Background:
LVO accounts for 40% of all ischemic strokes. While recent studies have shown that AI softwares have good accuracy in identifying LVOs on computed tomography angiograms (CTA), whether this improves workflow times and patient outcomes is still unclear.
Design/Methods:
PubMed, Embase, and Web of Science were systematically searched for observational and randomized controlled trials that compared initial radiological assessment assisted by AI softwares versus standard assessment of patients with acute LVO strokes. Results were pooled as mean differences for continuous outcomes, and odds ratio (OR) for dichotomous outcomes, along with 95% confidence intervals (CI). A random-effects model was used for all analyses.
Results:
A total of 8 studies comprising 984 patients were included, of whom 54.7% (538) had AI-assisted radiological assessment. AI consistently improved treatment times when compared to standard assessment, as evidenced by a mean reduction of 24.79 minutes in door-to-groin time (95% CI -42.85 to -6.73; p<0.01; Figure 1) and a reduction of 14.99 minutes in CTA to reperfusion (95% CI -28.45 to -1.53; p=0.03). Functional independence, defined as a modified Rankin scale 0-2, occurred more frequently in the AI-supported group when compared to the standard workflow (OR 1.58; 95% CI 1.01 to 2.47; p=0.04; Figure 2). Mortality, however, was similar among groups (OR 0.71; 95% CI 0.27 to 1.88; p=0.49).
Conclusions:
The incorporation of AI softwares for LVO detection in acute ischemic stroke enhanced workflow efficiency, reduced treatment duration, and improved rates of functional independence.