Nonlinear dynamical analysis of chaos, fractal complexity, and entropy in EEG signals during music therapy in ICU burn patients.

Music therapy (MT) is known to influence brain dynamics; however, its effects on nonlinear electroencephalogram (EEG) characteristics in clinical settings remain underexplored. This study investigated how MT modulates neural complexity and frequency-specific dynamics using nonlinear time-series analysis. EEG recordings from ICU burn patients were analyzed using Higuchi's Fractal Dimension (HFD), Sample Entropy (SE), Detrended Fluctuation Analysis (DFA), Multifractal DFA (MFDFA), Largest Lyapunov Exponent (LLE), and Permutation Entropy (PE). These features characterized changes in neural complexity, sensitivity to initial conditions, and fractal scaling behavior. Statistical comparisons and machine learning classifiers, validated using a rigorous Leave-One-Subject-Out framework, were employed to assess changes across pre-, during-, and post-MT phases. The analysis revealed that MT induced a global increase in geometric complexity and chaos, most prominently in broadband and theta activity, evidenced by higher SE, and LLE values. Concurrently, a reduction in long-range correlations was observed, indicated by a lower DFA scaling exponent. Contrary to the rigid dynamics often associated with stress, these changes suggest a transition toward 'dynamic relaxation,' characterized by restored neural flexibility and higher dimensional complexity. Machine learning analysis confirmed the robustness of these markers, with nonlinear features achieving high classification accuracy and significantly outperforming conventional spectral power features in differentiating therapeutic phases. These findings suggest that MT promotes a shift toward a more flexible dynamical regime that may benefit stress reduction and cognitive regulation, while highlighting nonlinear EEG features as promising objective biomarkers for therapeutic monitoring in critical care.