脑机接口控制外骨骼训练用于脊髓损伤下肢康复：一项初步随机临床试验

OBJECTIVE: This study aimed to evaluate the efficacy of brain-computer interface (BCI)-controlled exoskeleton training on lower-limb functional recovery, psychological outcomes, and neural plasticity in patients with spinal cord injury (SCI).

METHODS: We conducted a single-center, prospective, randomized, single-blind pilot trial (ChiCTR2300074503) including 21 patients with SCI. Participants were randomized to a BCI-exoskeleton group (B + E, n = 10) or an exoskeleton-only group (E, n = 11) for lower-limb training. Both groups received conventional rehabilitation plus 30 minutes of training, 6 days per week, for 4 weeks. The primary outcomes were Walking Index for Spinal Cord Injury II (WISCI II) scoring. Secondary outcomes included Lower Extremity Motor Score (LEMS), Spinal Cord Independence Measure version III (SCIM III), International Association of Neurorestoratology Spinal Cord Injury Functional Rating Scale (IANR-SCIFRS), 10-Meter Walk Test (10MWT), 6-Minute Walk Test (6MWT), and Hospital Anxiety and Depression Scale (HADS). Cortical plasticity was assessed by electroencephalography (EEG) and magnetic resonance imaging (MRI).

RESULTS: The B + E group showed a significant improvement in LEMS (p = 0.003), whereas both groups improved in IANR-SCIFRS (p < 0.05). The B + E group demonstrated significant within-group gains in walking speed (10MWT, p < 0.001) and endurance (6MWT, p = 0.031), although between-group differences were not significant. Compared with the E group, the B + E group had larger reductions in HADS scores (p = 0.003). EEG analyses revealed stronger μ/β desynchronization and increased network efficiency, whereas MRI showed no structural changes.

INTERPRETATION: BCI-controlled exoskeleton training enhanced motor function, walking performance, and depressive symptoms more than exoskeleton training alone, likely through cortical reorganization. Extended training may further consolidate these benefits, supporting BCI-exoskeleton integration as a promising rehabilitation strategy for SCI. ANN NEUROL 2026;99:1124-1138.