Stress Estimation in Elderly Oncology Patients Using Visual Wearable Representations and Multi-Instance Learning

2026-04-08 · Source: Takara TLDR - Daily AI Papers · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Internet of Things (IoT) & Connected Devices · Depth: Expert, quick

Summary

A study estimates perceived stress in elderly breast cancer patients within the CARDIOCARE cohort using multimodal wearable data from smartwatches and chest-worn ECG sensors. The research transforms physical activity and sleep data into heterogeneous visual representations, creating a weakly supervised setting where a single Perceived Stress Scale (PSS) score corresponds to multiple unlabeled data windows. A lightweight pretrained mixture-of-experts backbone, Tiny-BioMoE, embeds these representations into 192-dimensional vectors. These vectors are then aggregated using attention-based Multiple Instance Learning (MIL) to predict PSS scores at month 3 (M3) and month 6 (M6). Under leave-one-subject-out (LOSO) evaluation, the predictions showed moderate agreement with questionnaire scores, achieving R^2 values of 0.24 at M3 and 0.28 at M6, with global RMSE/MAE of 6.62/6.07 at M3 and 6.13/5.54 at M6.

Key takeaway

For AI Scientists developing health monitoring systems, this research demonstrates a viable approach to integrate wearable data for continuous stress estimation in oncology patients. You should consider adopting multimodal data fusion and Multiple Instance Learning techniques to address the challenges of weakly supervised settings and sparse ground truth labels in clinical applications. This method offers a pathway to enhance cardiotoxicity surveillance beyond traditional PROMs.

Key insights

Wearable data and MIL can estimate perceived stress in oncology patients, offering continuous cardiotoxicity surveillance.

Principles

• Multimodal wearable data improves stress estimation.
• Weakly supervised learning is effective for sparse labels.

Method

Transform wearable data into visual representations, embed with Tiny-BioMoE, then aggregate via attention-based Multiple Instance Learning to predict PSS scores.

In practice

• Integrate wearables for continuous patient monitoring.
• Apply Tiny-BioMoE for efficient embedding.

Topics

• Stress Estimation
• Elderly Oncology Patients
• Wearable Sensors
• Multi-Instance Learning
• Tiny-BioMoE

Best for: AI Scientist, Research Scientist

Related on AIssential

• Personalized and Context-Aware Transformer Models for Predicting Post-Intervention Physiological Responses from Wearable Sensor Data — Transformer models can predict personalized physiological responses to stress interventions using wearable sensor data.
• Multi-FRuGaL: Multimodal Flexible Redundancy-aware Decomposed Gated Learning for Cancer Diagnosis and Prognosis — Multi-FRuGaL robustly diagnoses cancer from incomplete multimodal medical data by separating redundant from complementary signals.
• End-to-End Machine Learning for Depressive State Classification via EEG and fNIRS — Biological signal analysis via EEG and fNIRS offers an objective framework for depressive state classification.
• Intervention-Based Self-Supervised Learning: A Causal Probe Paradigm for Remote Photoplethysmography — Intervention-based self-supervised learning can overcome correlation traps in rPPG by actively verifying physical signal hypotheses.
• 🔬 Training Transformers to solve 95% failure rate of Cancer Trials — Ron Alfa & Daniel Bear, Noetik — Improving cancer drug success requires advanced AI models for precise patient selection based on multimodal biological data.
• A New Privacy-First AI Predicts COVID Severity Using X-Rays and Medical Records — A privacy-first AI predicts COVID-119 severity using federated learning on X-rays and medical records.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.