P127 Disentangling Temporal and Amplitude-Driven Contributions to Signal Complexity
Sara Kamali¹,Fabiano Baroni¹, Pablo Varona¹
¹ Department of Computer Engineering, Autonomous University of Madrid, Madrid, Spain
*Email: sara.kamali@uam.es Introduction
Quantifying complexity in biomedical signals is crucial for physiological and pathological analysis. Entropy-based methods, like Shannon [1], approximate entropy [2], and sample entropy (SampEn) [3] quantify unpredictability. Some approaches, including increment-based methods [4, 5], capture entropies from amplitude variations. Existing methods, however, do not distinguish complexity derived from temporal dynamics versus amplitude fluctuations. This limitation restricts insights into the dynamical evolution of signals We introduce Extrema-Segmented Entropy (ExSEnt), an entropy-based framework that independently analyzes temporal and amplitude components, enhancing understanding of underlying dynamics. Methods We segmented the time series based on extrema, each segment starts at the data point after the current extremum and ends at the next extremum. Two key features were extracted per segment: duration, representing the temporal length, and net amplitude, reflecting the overall signal variation. We then computed SampEn for each feature separately, as well as their joint bivariant entropy, to assess whether they provide independent or correlated information. This approach helps determine whether complexity arises primarily from temporal dynamics or amplitude variations. Our method enhances the understanding of how different factors drive signal complexity. Results Application of ExSEnt on synthetic data revealed the ability of the metrics to distin- guish between different random signals, i.e., Gaussian noise, pink noise, and Brownian motion. We also evaluated the complexity of well-known dynamical systems, such as the Rulkov neuron model, where ExSEnt successfully differentiated between different dy- namical regimes. Evaluation of electromyography (EMG) signals during a motor task revealed that movement intervals exhibit lower amplitude complexity but relatively sta- ble temporal entropy compared to the baseline. A strong linear correlation was observed between amplitude ExSEnt and joint ExSEnt, suggesting that amplitude variations are the primary contributors to the joint amplitude-temporal EMG complexity. Discussion The ExSEnt framework offers a precise and systematic approach to quantifying tempo- ral and amplitude-driven contributions to complexity, providing a novel perspective for biomedical and neuronal signal analysis. Applying ExSEnt to neural data demonstrates its potential to reveal hidden dependencies between duration and amplitude fluctuations, pro- viding a detailed complexity profile. This approach aids in quantifying dynamic changes and identifying complexity sources in neural disorders and physiological states.
Acknowledgements Work funded by PID2024-155923NB-I00, CPP2023-010818, and PID2021-122347NB-I00. References [1] https://doi.org/10.1002/j.1538-7305.1948.tb01338.x [2] https://doi.org/10.1073/pnas.88.6.2297 [3] https://doi.org/10.1016/S0076-6879(04)84011-4 [4] https://doi.org/10.3390/e20030210 [5] https://doi.org/10.3390/e18010022
