FuzzyEn Compared to SampEn for Evaluation of Dynamic Complexity

Fuzzy entropy (FuzzyEn) theoretically outperforms sample entropy (SampEn) in quantifying dynamic complexity; however, its anti-noise robustness remains insufficiently validated. This study compares SampEn and FuzzyEn using model simulations and real-world depression electroencephalogram (EEG) signals. The two entropy metrics are first compared using chaotic time series generated from the logistic and two-dimensional Henon maps, with additive white Gaussian noise (SNRs ranging from 1 to 10 dB). Surrogate data analysis reveals that SampEn maintains stronger nonlinear detection performance (lower than the 2.5th percentiles of surrogate data) under noisy conditions. EEG data from 46 depressed patients and 75 healthy subjects are employed to evaluate SampEn and FuzzyEn, with statistical differences corrected by the Bonferroni method. Depressed patients showed significantly increased EEG complexity in the occipital lobe (PO7, PO5, PO3 channels) under visual stimulation (p < 0.01). SampEn outperforms FuzzyEn in alpha-band feature detection, especially the low-alpha sub-band (8–10 Hz, p < 0.001). In summary, this study verifies that SampEn may be more suitable than FuzzyEn under noise conditions, thus providing valuable insights for real-world signal analysis and methodological guidance for developing EEG biomarkers of depression.