Gonzalez Nunez, J. G. | bioRxiv Science

bioRxiv2026

Predictive Feature Engineering for Stress Detection using Physiological Signals, A Comparative Study

This paper presents a two-stage pipeline for implicit feature engineering in time series-based physiological stress detection using electrodermal activity (EDA) signals. In the first stage, we forecast three descriptive statistics of future EDA signals over short horizons (3, 5, and 10 seconds) based on a 60-second context window. In the second stage, a lightweight linear classifier detects stress from these predicted statistics. We evaluate three forecasting architectures spanning the domain expertise spectrum: a domain-specific bidirectional long short-term memory (BiLSTM) recurrent neural network, zero-shot and fine-tuned variants of Amazon Chronos T5 time series foundation model, and the Tabular Prior-data Fitted Network (TabPFN) applied to engineered physiological features. Experiments on the publicly available Wearable Stress and Affect Detection (WESAD) dataset, comprising chest-worn multimodal physiological signals from 15 subjects under baseline and stress conditions, use subject-independent 5-fold cross-validation and show that the domain-specific BiLSTM and TabPFN achieve comparable classification performance, with mean area under the receiver operating characteristic curve (AUC) values of 0.859-0.882 and 0.863-0.883 respectively. Both remain well ahead of the Chronos variants, which yield 0.629-0.777. Chronos models quickly reach performance saturation regardless of training depth, highlighting challenges in tokenizing continuous physiological time series. The proposed approach advances implicit feature engineering for wearable stress monitoring by leveraging forecasting as a powerful inductive bias, thereby improving robustness and providing insights into the limitations of the foundation model for physiological signals.

bioengineering↗

bioRxiv2025

Comparative Analysis of Diffusion Models for Enhancing Alzheimer's Disease Classification

Early and accurate detection of Alzheimers disease (AD) is vital for timely intervention and better patient outcomes. However, training machine learning (ML) models for this purpose is challenging due to the limited medical images available and the imbalance of classes. The size and quality of the training dataset directly affect model performance. Recent advances in diffusion models address this limitation by generating synthetic images from a small sample of real images. In this work, we adapted two diffusion models and trained VGG16 and ConvNeXt classification models for AD classification. The first diffusion model was a Denoising Diffusion Probabilistic Model (DDPM) with a U-Net architecture, and the second was a U-KAN framework that integrates Kolmogorov-Arnold Networks (KANs) with the U-Net. Both models were fine-tuned to generate MRI scans of AD or Late Mild Cognitive Impairment (LMCI). We conducted a comparative analysis to assess the reliability and usefulness of these synthetic images for training classification models. The best metrics achieved by the classification models using synthetic images for the AD class were precision of 96%, recall of 83%, F1 score of 87%, and AUC of 0.88. For the LMCI class, the best values were precision of 78%, recall of 88%, F1 score of 82%, and AUC of 0.88. They both demonstrated noticeable improvement from the baseline trained only on the original images.

bioinformatics↗