Related Experiment Video
Updated: May 23, 2025
11:21
Methodology for Establishing a Community-Wide Life Laboratory for Capturing Unobtrusive and Continuous Remote Activity and Health Data
Published on: July 27, 2018
8.2K
Development and Feasibility Study of HOPE Model for Prediction of Depression Among Older Adults Using Wi-Fi-based
Shayan Nejadshamsi1,2,3, Vania Karami1,2,3, Negar Ghourchian4
1Mila-Quebec Artificial Intelligence Institute, Montreal, QC, Canada.
JMIR Aging
|March 7, 2025
Summary
This study shows that nonintrusive Wi-Fi sensors can detect depression in older adults. Sleep patterns and frailty were key indicators identified by the HOPE machine learning model.
Area of Science:
- Gerontology
- Machine Learning
- Digital Health
Background:
- Depression significantly impacts quality of life, necessitating early detection.
- Traditional depression classification methods often use intrusive wearables and large datasets.
- Existing studies lack explainability in their single-stage classifiers.
Purpose of the Study:
- To assess the feasibility of classifying depression using nonintrusive Wi-Fi sensor data.
- To evaluate a novel machine learning model (HOPE) for depression prediction in older adults.
- To conduct an explainability analysis to identify key depression indicators.
Main Methods:
- Recruited adults aged 65+ for a 6-month data collection period.
- Utilized nonintrusive Wi-Fi sensors for activity and sleep data, alongside clinical scales.
- Developed and evaluated the HOPE model with feature selection, dimensionality reduction, and classification stages.
Main Results:
- The most accurate model achieved 87.5% accuracy, 90% sensitivity, and 88.3% precision.
- Explainability analysis identified sleep duration, sleep interruptions, and frailty scale as key features.
- One participant was classified as having depression out of four remaining participants.
Conclusions:
- Wi-Fi sensor data and the HOPE model show feasibility for depression classification, even with small sample sizes.
- The nonintrusive approach is promising for remote health monitoring in older adults.
- Further research with larger cohorts is recommended to validate these findings.
