Walking-in-Place Characteristics-Based Geriatric Assessment Using Deep Convolutional Neural Networks.

Area of Science:

• Gerontology and Biomedical Engineering
• Focuses on aging populations and developing innovative health monitoring technologies.

Background:

• Global population aging presents challenges for healthcare systems.
• Need for accessible, in-home monitoring solutions for elderly health conditions like frailty and cognitive dysfunction.

Purpose of the Study:

• To develop a simple, reliable in-home method for monitoring elderly frailty and cognitive dysfunction.
• To utilize walking-in-place characteristics for health assessment.

Main Methods:

• Fifty-four elderly participants (≥65 years) were assessed using the FRAIL scale and mini-mental state examination.
• Inertial measurement units recorded lower body movement during a 20-second walk-in-place task.
• Time-frequency analysis generated walking-in-place spectrograms for deep convolutional neural network classification.

Main Results:

• Classifiers achieved 94.63% accuracy in distinguishing robust from non-robust elderly groups.
• Classifiers achieved 97.59% accuracy in differentiating cognitive impairment from non-cognitive impairment groups.
• Walking-in-place spectrograms proved effective indicators for frailty and cognitive status.

Conclusions:

• Walking-in-place spectrogram analysis offers a feasible method for in-home monitoring of elderly frailty and cognitive function.
• This approach requires minimal space and equipment, making it suitable for daily use.
• The findings support the use of gait analysis for proactive elderly care.