2025 American Academy of Neurology Abstract Website

Valentina Velasco-Muñoz¹, Andrés Ricaurte-Fajardo², Kiara Torres-Tello¹, Salomón Páez-García¹, Gabriel Castellanos-Castañeda³, Hernando Santamaría-García⁴, Juan Carlos Acevedo-González⁵
¹Medical School, Pontificia Universidad Javeriana, Bogotá, Colombia, ²Departament of Neuroscience, Division of Neurology, Pontificia Universidad Javeriana, Bogotá, Colombia, ³Department of Physiological Sciences, Pontificia Universidad Javeriana, Bogotá, Colombia, ⁴Department of Psychiatry, Pontificia Universidad Javeriana, Bogotá, Colombia, ⁵Department of Neuroscience, Division of Neurosurgery, Pontificia Universidad Javeriana, Bogotá, Colombia

Objective:

This systematic review aims to evaluate the role of artificial intelligence (AI) in improving the diagnosis of patients with disorders of consciousness (DOC), such as coma, vegetative state, and minimally conscious state, and identify how AI can enhance diagnostic accuracy in this challenging population.

Background:

Disorders of consciousness (DOC), including coma, vegetative state (VS), and minimally conscious state (MCS), pose significant diagnostic challenges in neurology. Recent advances in artificial intelligence (AI), particularly in neuroimaging analysis, offer promising solutions to enhance diagnostic accuracy and improve clinical decision-making in this complex field.

Design/Methods:

A systematic search was performed in PubMed, Embase, and Scopus for studies using AI tools, like machine learning, to aid in diagnosing disorders of consciousness (DOC). Eligible studies were articles utilizing neuroimaging techniques and reporting diagnostic accuracy metrics. Excluded were non-AI studies, those focused on other neurological conditions, case reports, reviews and books. Reviewers independently screened titles and abstracts, and full-text articles were evaluated for eligibility, with any discrepancies resolved through discussion.

Results:

AI-based methods, including machine learning models such as SVM, XGBoost, and DeepDOC, demonstrated high accuracy in diagnosing DOC states. Techniques like EEG microstate analysis, P300 signal detection, and fractal dimension analysis contributed to improved classification of DOC states and prognostic outcomes. AI-enhanced neuroimaging (EEG, fMRI, PET) facilitated the detection of covert consciousness (CMD) in previously undiagnosed patients, and early EEG measures showed potential for predicting long-term recovery outcomes.

Conclusions:

AI, particularly machine learning in combination with neuroimaging, holds significant promise for improving diagnostic accuracy and prognostic assessment in DOC patients.While current studies highlight its promise, further research with larger, more diverse cohorts and longer follow-up periods is essential to validate and expand its clinical application.