Related Experiment Video
Updated: Jan 17, 2026
08:56
Clinical Assessment of Spatiotemporal Gait Parameters in Patients and Older Adults
Published on: November 7, 2014
14.3K
Retrospective Frailty Assessment in Older Adults Using Inertial Measurement Unit-Based Deep Learning on Gait
Julius Griškevičius1, Kristina Daunoravičienė1, Liudvikas Petrauskas1
1Department of Biomechanical Engineering, Vilnius Gediminas Technical University, LT-10105 Vilnius, Lithuania.
Sensors (Basel, Switzerland)
|September 19, 2025
Summary
This study shows deep learning can accurately detect frailty in older adults using wearable sensors. Analyzing gait spectrograms with a convolutional neural network offers a promising objective frailty assessment tool.
Area of Science:
- Gerontology
- Biomedical Engineering
- Artificial Intelligence
Background:
- Frailty is a significant geriatric syndrome increasing risks of falls, disability, and mortality.
- Early and accurate frailty detection is crucial for timely interventions and improved health outcomes in older adults.
- Current frailty assessment methods often rely on subjective clinical judgment, leading to potential inconsistencies.
Purpose of the Study:
- To investigate the efficacy of deep learning models in classifying frailty using inertial measurement unit (IMU) data.
- To evaluate the performance of a convolutional neural network (CNN) trained on gait spectrograms for objective frailty assessment.
Main Methods:
- Retrospective analysis of an existing IMU dataset collected during gait analysis.
- Placement of six IMUs on lower extremity segments to capture motion data during walking.
- Conversion of raw accelerometer and gyroscope signals into time-frequency spectrograms.
- Training a CNN model exclusively on these raw IMU-derived spectrograms.
Main Results:
- The CNN model achieved 71.4% subject-wise accuracy in distinguishing between Frail, PreFrail, and NoFrail groups.
- Minimal data preprocessing did not enhance the model's performance, indicating the salience of raw time-frequency gait cues.
- The findings highlight the potential of raw spectrograms for capturing essential gait characteristics.
Conclusions:
- Wearable sensor technology combined with deep learning presents a robust and objective method for frailty assessment.
- This approach holds significant potential for integration into clinical practice and remote health monitoring systems.
- Objective frailty classification can facilitate personalized interventions and improve geriatric care.
