Quantitative electroencephalography (QEEG) has emerged as a promising tool for detecting Alzheimer’s disease (AD). Among QEEG measures, Multiscale Fuzzy Entropy (MFE) shows great potential in identifying AD-related changes in EEG complexity. However, MFE is intrinsically linked to signal amplitude, which can vary substantially among EEG systems, and this hinders the adoption of this metric for AD detection. To overcome this issue, this study investigates different preprocessing strategies to make the calculation of MFE less dependent on the specific amplitude characteristics of the EEG signals at hand. This contributes to generalizing and making more robust the adoption of MFE for AD detection. To demonstrate the robustness of the proposed preprocessing methods, binary classification tasks with Support Vector Machines (SVMs), Random Forest (RF), and K-Nearest Neighbor (KNN) classifiers are used. Performance metrics, such as classification accuracy and Matthews Correlation Coefficient (MCC), are employed to assess the results. The methodology is validated on two public EEG datasets. Results show that amplitude transformation, particularly normalization, significantly enhances AD detection, achieving mean classification accuracy values exceeding 80% with an uncertainty of 10% across all classifiers. These results highlight the importance of preprocessing in improving the accuracy and the reliability of EEG-based AD diagnostic tools, offering potential advancements in patient management and treatment planning.
This study investigates the effectiveness of amplitude transformation in enhancing the performance and robustness of Multiscale Fuzzy Entropy for Alzheimer’s disease detection using electroencephalography signals. Multiscale Fuzzy Entropy is a complexity measure particularly sensitive to intra- and inter-subject variations in signal amplitude, as well as the selection of key parameters such as embedding dimension (m) and similarity criterion (r), which often result in inconsistent outcomes when applied to multivariate data, such as electroencephalography signals. To address these challenges and to generalize the possibility of adopting Multiscale Fuzzy Entropy as a diagnostic tool for Alzheimer’s disease, this research explores amplitude transformation preprocessing on electroencephalography signals in Multiscale Fuzzy Entropy calculation across varying parameters. The statistical analysis of the obtained results demonstrates that amplitude transformation preprocessing significantly enhances Multiscale Fuzzy Entropy’s ability to detect Alzheimer’s disease, achieving higher and more consistent significant comparison percentages, with an average of 73.2% across all parameter combinations, compared with only one raw data combination exceeding 65%. Clustering analysis corroborates these findings, showing that amplitude transformation improves the differentiation between Alzheimer’s disease patients and healthy subjects. These results highlight the potential of amplitude transformation to stabilize Multiscale Fuzzy Entropy performance, making it a more reliable tool for early Alzheimer’s disease detection.
1 April 2026 in IEEE Transactions on Biomedical Engineering
Signal complexity analysis plays a crucial role in biomedical research, particularly in electroencephalography (EEG), for early disease diagnosis and cognitive monitoring. However, traditional entropy-basedmethods lack robustness, suffer from limitations such as sensitivity to noise, and fail to capture themulti-frequency structure of brain signals. To address these challenges, this study introduces Multivariate Multiscale Multi-Frequency Entropy (M3FrEn), a novel complexity metric that simultaneously incorporates multiscale dynamics, multichannel dependencies, and multi-frequency structure into a unified entropy-based framework. M3FrEn was evaluated using both simulated signals and real EEG data from 52 healthy subjects and 52 patients with Alzheimer's Disease (AD). Compared to existing entropy-based metrics, simulated tests demonstrated that M3FrEn reliably discriminates signals with varying complexity levels and maintains stability across embedding dimensions and data lengths. Applied to real EEG data, M3FrEn identified statistically significant differences ($p_{value} < 0.05$) between AD and control groups across multiple brain regions and frequency bands. Compared to benchmark methods, M3FrEn yielded more robust and consistent group separation. In addition, binary classification based on M3FrEn features achieved $85\%$ accuracy, with $100\%$ sensitivity and $70\%$ specificity. These results demonstrate that M3FrEn offers a powerful tool for EEG-based complexity analysis and holds promise for clinical applications in early-stage neurological disease detection.
Machine Learning (ML) techniques are increasingly used in measurement science to develop novel measurement models. Such ML-based measurement models are particularly useful when conventional ones are unavailable, enabling measurement operations that would otherwise be impractical. However, the resulting relationship between the measurand and the input quantities is not directly selected by the practitioner; rather, it is inherently determined by a data-driven training process. This gives rise to an additional uncertainty component (denoted in previous work as measurement model uncertainty) which, when adherence to the Joint Committee For Guides in Metrology (JCGM) is required, shall be negligible with respect to the component associated with the measurement of the physical input quantities (denoted in previous work as physical quantity uncertainty). When this condition cannot be met, a direct application of the JCGM framework may understate the measurement uncertainty, thereby yielding coverage intervals that are not representative of the values that can be reasonably attributed to the measurand. Accordingly, to address the limitations of a direct JCGM application while remaining consistent with its framework, this work proposes a JCGM-compliant method that leverages joint probability distributions, sampled via nested Monte Carlo simulations, to obtain a fully fledged measurement result for ML-assisted measurements. Two illustrative case studies based on deterministic ML regressors demonstrate the applicability of the proposed method and show that the resulting coverage intervals explicitly account for the variability introduced by the ML model, thereby avoiding any understatement of the reported measurement uncertainty.