Neurology. Education · 2026
Education Research: Validation of an AI-Powered Educational Platform for Neuroanatomical Localization and Stroke Syndrome Recognition: A Multicenter Study.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Fagundes Thales Pardini, de Andrade João Brainer Clares, Mota Alice Marilia Silva Abreu, Nishimura Victor Dyego Mena Romeiro, Soga Felipe Hideki, Camilo Millene
Аннотация
BACKGROUND AND OBJECTIVES: Neuroanatomical localization is foundational to stroke practice but remains difficult across training levels. We evaluated whether LouiS, an AI educational platform using retrieval-augmented generation over a curated stroke-syndrome knowledge base, improves the proportion of correct responses for neuroanatomical localization and stroke-syndrome recognition, with a focus on its potential as a teaching aid.
METHODS: This two-phase study recruited 80 participants from 3 Brazilian university-affiliated centers. Phase 1 used web-app allocation with browser-fingerprint persistence to keep each participant in the same test condition, with or without LouiS, across repeated access. Phase 2 used a within-participant paired design in which 62 participants evaluated case-matched vignettes first without and then with LouiS. Participant-level scores used the actual number of scorable responses in each condition. The allocation protocol required exposure to higher-difficulty vignettes and supplemented them with randomly selected cases. The primary outcome was the proportion of correct responses against a consensus reference standard.
RESULTS: Across the pooled response-wise dataset including phase 1 and available phase 2 responses, correct responses increased from 63.4% without LouiS to 71.9% with LouiS (absolute difference 8.5 percentage points; 95% CI 2.79-14.20; generalized estimating equation OR 1.36, 95% CI 1.11-1.67; p = 0.003). In phase 2, median participant-level proportion correct increased from 62.1% [IQR 60.0-80.0] without LouiS to 80.0% [IQR 60.0-100.0] with LouiS, a mean within-participant gain of 7.9 percentage points (95% CI 4.10-11.75; Wilcoxon W = 23.0, p = 0.0002; Cohen d = 0.53). In paired response-wise analysis, 30 responses changed to correct and 5 to incorrect (McNemar χ2 = 16.46, p < 0.0001). Deviations clustered on differential-diagnostic pairs.
DISCUSSION: Use of LouiS was associated with improved performance in neuroanatomical localization and stroke-syndrome recognition during case-based testing. The findings support LouiS as an educational aid for neuroanatomical reasoning and continuing medical education. Knowledge retention, behavior change, and real-world clinical outcomes require separate prospective evaluation.
METHODS: This two-phase study recruited 80 participants from 3 Brazilian university-affiliated centers. Phase 1 used web-app allocation with browser-fingerprint persistence to keep each participant in the same test condition, with or without LouiS, across repeated access. Phase 2 used a within-participant paired design in which 62 participants evaluated case-matched vignettes first without and then with LouiS. Participant-level scores used the actual number of scorable responses in each condition. The allocation protocol required exposure to higher-difficulty vignettes and supplemented them with randomly selected cases. The primary outcome was the proportion of correct responses against a consensus reference standard.
RESULTS: Across the pooled response-wise dataset including phase 1 and available phase 2 responses, correct responses increased from 63.4% without LouiS to 71.9% with LouiS (absolute difference 8.5 percentage points; 95% CI 2.79-14.20; generalized estimating equation OR 1.36, 95% CI 1.11-1.67; p = 0.003). In phase 2, median participant-level proportion correct increased from 62.1% [IQR 60.0-80.0] without LouiS to 80.0% [IQR 60.0-100.0] with LouiS, a mean within-participant gain of 7.9 percentage points (95% CI 4.10-11.75; Wilcoxon W = 23.0, p = 0.0002; Cohen d = 0.53). In paired response-wise analysis, 30 responses changed to correct and 5 to incorrect (McNemar χ2 = 16.46, p < 0.0001). Deviations clustered on differential-diagnostic pairs.
DISCUSSION: Use of LouiS was associated with improved performance in neuroanatomical localization and stroke-syndrome recognition during case-based testing. The findings support LouiS as an educational aid for neuroanatomical reasoning and continuing medical education. Knowledge retention, behavior change, and real-world clinical outcomes require separate prospective evaluation.