Multimodal Stress Monitoring Using Deep Learning for Telemedicine and Telehealth Applications

Detection and monitoring of stress are crucial in telemedicine applications, as stress can lead to critical health issues. Timely intervention enhances patient monitoring and aids in the early identification of potential mental health conditions. Traditionally, telemedicine platforms rely on questionnaires and psychological assessments to evaluate stress levels and identify underlying causes. Though there have been significant advancements in integrating wearable devices with telemedicine platforms, challenges related to data accuracy and high costs still persist. To address these challenges, we propose a novel architecture utilizing multiple physiological parameters. The dataset used for training and testing is sourced from the A Database for Emotion Analysis using Physiological Signals (DEAP), a multi-modal dataset designed for the categorization of different emotions. The dataset includes various physiological and EEG signal recordings from 32 participants. The proposed architecture utilized Electroencephalography (EEG) and other physiological parameters to classify features using one-dimensional Convolutional Neural Networks (CNNs) combined with fully connected layers. We evaluated the performance across varying input sequence lengths, and the obtained results indicate that shorter segment lengths achieve higher metrics in terms of accuracy and F1-Score. The model achieved a maximum accuracy of 88.27% to distinguish stress with non-stress signals with a 10-second input sequence.