Evaluation of Scale-Invariance In Physiological Signals By Means Of Balanced Estimation Of Diffusion Entropy

By means of the concept of balanced estimation of diffusion entropy we evaluate reliable scale-invariance embedded in different sleep stages and stride records. Segments corresponding to Wake, light sleep, REM, and deep sleep stages are extracted from long-term EEG signals. For each stage the scaling value distributes in a considerable wide range, which tell us that the scaling behavior is subject- and sleep cycle- dependent. The average of the scaling exponent values for wake segments is almost the same with that for REM segments ($\sim 0.8$). Wake and REM stages have significant high value of average scaling exponent, compared with that for light sleep stages ($\sim 0.7$). For the stride series, the original diffusion entropy (DE) and balanced estimation of diffusion entropy (BEDE) give almost the same results for de-trended series. Evolutions of local scaling invariance show that the physiological states change abruptly, though in the experiments great efforts have been done to keep conditions unchanged. Global behaviors of a single physiological signal may lose rich information on physiological states. Methodologically, BEDE can evaluate with considerable precision scale-invariance in very short time series ($\sim 10^2$), while the original DE method sometimes may underestimate scale-invariance exponents or even fail in detecting scale-invariant behavior. The BEDE method is sensitive to trends in time series. Existence of trend may leads to a unreasonable high value of scaling exponent, and consequent mistake conclusions.

💡 Research Summary

The paper introduces a novel statistical tool called Balanced Estimation of Diffusion Entropy (BEDE) and demonstrates its superiority over the conventional Diffusion Entropy (DE) method for detecting scale‑invariant behavior in physiological time series. DE estimates a scaling exponent α by measuring how the Shannon entropy of a diffusion process grows with time. However, when the data length is short (on the order of 10² points), DE suffers from sampling bias, boundary effects, and often underestimates α or fails to detect scaling altogether. BEDE addresses these issues by re‑weighting the probability estimates for each diffusion window and correcting the bias in the logarithmic probability term, thereby delivering reliable α values even for very short segments.

Two experimental datasets were analyzed. The first consisted of long‑term electroencephalogram (EEG) recordings from sleep studies. Segments corresponding to Wake, Light‑Sleep, REM, and Deep‑Sleep were extracted, each lasting several minutes. The second dataset comprised stride‑length time series recorded from subjects walking at a steady pace. For each segment, both DE and BEDE were applied to compute the scaling exponent.

In the EEG analysis, BEDE revealed that α values are broadly distributed between 0.7 and 0.85 across all sleep stages, indicating pronounced subject‑ and cycle‑dependence. Wake and REM periods showed the highest average α (≈0.80), suggesting stronger long‑range correlations, whereas Light‑Sleep exhibited a lower average α (≈0.70). Deep‑Sleep produced intermediate values (≈0.75) but with considerable inter‑subject variability. These findings demonstrate that a single global exponent is insufficient to characterize the dynamical state of the brain; instead, a time‑resolved, stage‑specific scaling analysis captures the nuanced transitions between physiological states.

For the stride‑length series, the original DE and BEDE gave nearly identical results (α≈0.90) after detrending the data. However, when a subtle trend remained in the raw series, DE tended to underestimate α or missed scaling, while BEDE produced an unrealistically high exponent (>1.2). This sensitivity to trends underscores the necessity of proper preprocessing (e.g., detrending) before applying BEDE, as the method will otherwise amplify spurious correlations.

Methodologically, BEDE offers three key advantages: (1) accurate exponent estimation for short records (~10² points), enabling analysis of brief physiological events; (2) detection of weak, long‑range correlations that DE may overlook; and (3) heightened responsiveness to trends, which can be leveraged as a diagnostic check but also demands careful data cleaning. The authors argue that these properties make BEDE a powerful tool for extracting dynamic scaling biomarkers from physiological signals, with potential applications in sleep staging, fatigue monitoring, and early detection of neurological disorders.

The study concludes that physiological signals possess a rich, time‑varying multiscale structure that cannot be captured by a single static exponent. BEDE provides a robust framework for quantifying this structure, especially when data are limited in length or when rapid state changes occur. Future work is suggested to extend BEDE to other biosignals such as heart‑rate variability and electromyography, and to integrate the extracted scaling features into machine‑learning classifiers for real‑time health monitoring.