Prospective validation protocol for artificial intelligence-assisted standardization of Coma Recovery Scale-Revised scores using electroencephalography and clinical data in disorders of consciousness.

BACKGROUND: Disorders of consciousness (DoC) after severe acquired brain injury are difficult to diagnose at the bedside. The Coma Recovery Scale-Revised (CRS-R) is the behavioral reference standard, but raw scores can be affected by arousal fluctuation, motor impairment, sensory deficits, sedation, language disturbance, and local assessment conditions. OBJECTIVE: This protocol describes a prospective validation study of a multimodal artificial intelligence (AI) system designed to standardize CRS-R scores by integrating routine electroencephalography (EEG), bedside CRS-R subscale scores, and selected clinical contextual variables against an adjudicated standardized reference. METHODS AND ANALYSIS: The study includes retrospective model development followed by prospective, blinded clinical validation at Guangdong Second Traditional Chinese Medicine Hospital. The locked model will use subscale-specific ordinal prediction modules and a rule-constrained integration layer to generate AI-standardized CRS-R subscale scores, total scores, diagnostic categories, uncertainty estimates, and prespecified case-level explanations. Prospective participants will undergo two independent bedside CRS-R assessments and routine scalp EEG within target and maximum assessment windows. A blinded expert committee will generate the reference CRS-R score using paired behavioral records, available videos, nursing observations, medication and confounder records. The primary endpoint is the participant-level reduction in absolute CRS-R total-score error after AI-assisted standardization relative to the adjudicated reference. Secondary endpoints include subscale agreement, diagnostic-category agreement, inter-rater variability, uncertainty-flag performance, workflow feasibility, and exploratory day-14 consciousness improvement. ETHICS AND DISSEMINATION: The protocol was approved by the Ethics Committee of Guangdong Second Traditional Chinese Medicine Hospital (approval number Z202603-004-01; review date March 25, 2026; approval validity March 27, 2026 to March 26, 2027). Results will be disseminated through peer-reviewed publication and academic presentations. CLINICAL TRIAL REGISTRATION: https://www.chictr.org.cn, Chinese Clinical Trial Registry, ChiCTR2600124984.