Stress Estimation in Elderly Oncology Patients Using Visual Wearable Representations and Multi-Instance Learning
Summary: arXiv:2604.06990v1 Announce Type: cross
Abstract: Psychological stress is clinically relevant in cardio-oncology, yet it is typically assessed only through patient-reported outcome measures (PROMs) and is rarely integrated into continuous cardiotoxicity surveillance. We estimate perceived stress in an elderly, multicenter breast cancer cohort (CARDIOCARE) using multimodal wearable data from a smartwatch (physical activity and sleep) and a chest-worn ECG sensor.
Introduction
Psychological stress plays a significant role in the health outcomes of elderly oncology patients. In particular, individuals undergoing treatment for breast cancer may experience elevated levels of stress, which can adversely affect their overall well-being and treatment outcomes. Traditional methods for assessing stress, such as patient-reported outcome measures (PROMs), often fail to provide continuous monitoring and can overlook key physiological indicators of stress.
Methodology
To address this gap, the study employs a novel approach utilizing multimodal wearable technology. Specifically, data collected from smartwatches and chest-worn ECG sensors is analyzed to estimate perceived stress levels among elderly patients in the CARDIOCARE cohort. The wearable streams are transformed into heterogeneous visual representations, allowing for a weakly supervised learning environment.
Data Processing and Model Training
The research utilizes a lightweight pretrained mixture-of-experts backbone known as Tiny-BioMoE. This model processes the visual representations, embedding them into 192-dimensional vectors. The vectors are then aggregated using attention-based multiple instance learning (MIL) to predict perceived stress scale (PSS) scores at two key time intervals: month 3 (M3) and month 6 (M6).
Evaluation Metrics
The predictions from the model were evaluated using a leave-one-subject-out (LOSO) methodology. The findings reveal a moderate agreement between the predicted scores and the actual questionnaire scores. The details of the evaluation metrics are as follows:
- M3: R² = 0.24, Pearson r = 0.42, Spearman rho = 0.48
- M6: R² = 0.28, Pearson r = 0.49, Spearman rho = 0.52
- Global RMSE/MAE: M3: 6.62/6.07, M6: 6.13/5.54
Conclusion
The results of this study indicate that wearable technology can provide valuable insights into the psychological stress experienced by elderly oncology patients. By leveraging multimodal data and advanced machine learning techniques, healthcare providers may be able to enhance stress monitoring and improve patient outcomes. Future research should focus on refining these models and exploring their application across diverse patient populations in oncology.
Implications for Future Research
This innovative approach highlights the necessity for integrating continuous monitoring of psychological stress into clinical practice. Further studies are needed to validate these findings and expand the application of wearable technology in oncology.