Geriatric depression and anxiety screening via deep learning using activity tracking and sleep data.

Area of Science:

• Geriatric psychiatry
• Computational neuroscience
• Digital health

Background:

• Geriatric depression and anxiety are common mood disorders linked to dementia onset.
• Current diagnostic methods, including self-reports and clinical assessments, can be uncomfortable, prone to misreporting, time-consuming, and costly.
• There is a need for systematic, cost-effective approaches for diagnosing these conditions in older adults.

Purpose of the Study:

• To investigate the feasibility of training an end-to-end deep learning (DL) model using time-series activity tracking and sleep data.
• To identify comorbid depression and anxiety in older adults by directly inputting data from consumer-grade wrist-worn activity trackers.
• To develop a novel approach for mental health condition identification in geriatric populations.

Main Methods:

• Utilized time-series data from wrist-worn activity trackers, including step counts and sleep stages.
• Incorporated minimal depression and anxiety assessment scores as non-time-series data for model input.
• Applied various deep learning models (CNN, LSTM, ResNet) for mixed-input data processing and multi-label classification.

Main Results:

• Achieved significant results in multi-label classification of depression and anxiety with a Hamming loss of 0.0946 using a Residual Network (ResNet).
• Optimized time-series data processing through the comparison of hyper-parameter performance and development of diverse DL models.
• Demonstrated the effectiveness of a mixed-input DL model based on activity tracking data.

Conclusions:

• This study is the first to develop a mixed-input DL model using activity tracking data for identifying late-life depression and anxiety.
• Findings highlight the potential of consumer-grade activity trackers and DL models for improving mental health condition identification in older adults.
• Established a multi-label classification framework for complex depression and anxiety symptoms in the geriatric population.