Identifying Fatigue Indicators Using Gait Variability Measures: A Longitudinal Study on Elderly Brisk Walking

Guoxin Zhang, Ivy Kwan-Kei Wong, Tony Lin-Wei Chen, Tommy Tung-Ho Hong, Duo Wai-Chi Wong, Yinghu Peng, Fei Yan, Yan Wang, Qitao Tan, Ming Zhang2020 · DOI

TL;DR

This study uses a single inertial measurement unit (IMU) placed on the posterior heel to identify fatigue in elderly individuals during prolonged brisk walking. The research demonstrates that changes in gait variability, particularly in the medial-lateral direction, can serve as reliable indicators of fatigue, paving the way for wearable fatigue-detecting sensors for fall prevention.

What Problem Does This Paper Address?

As people age, their risk of falls increases significantly due to declines in muscle strength, coordination, and balance. Physical activity, like brisk walking, is crucial for maintaining health and reducing fall risk in the elderly. However, fatigue during exercise can exacerbate instability and increase the fear of falling, potentially leading to reduced physical activity. Current methods for detecting fatigue during walking often involve separate testing trials, which might not accurately capture the state of fatigue induced by the activity itself. This paper aims to address this by developing a non-intrusive, real-time method to identify fatigue during a continuous, fatigue-inducing brisk walking task, using a single IMU. The ultimate goal is to create wearable sensors that can alert elderly individuals to their fatigue levels, thereby preventing falls and encouraging continued physical activity.

Methods

The study employed a longitudinal design where 18 elderly participants (average age 60.4 years) underwent a 60-minute brisk walking trial on a treadmill. A single inertial measurement unit (IMU) was attached to the posterior heel of the dominant foot. This IMU captured data on acceleration, angular velocity, and orientation angles along the anterior-posterior, medial-lateral, and vertical axes. Data was collected at three time points: baseline, 30 minutes, and 60 minutes into the walk.

The IMU data was processed to extract ten kinematic parameters for each gait cycle, including median absolute deviation (MAD), kurtosis, skewness, root mean square (RMS), variance, maximum/minimum absolute value, amplitude range, and energy consumption (EC). To account for individual differences, these parameters were normalized by dividing the 30-minute and 60-minute values by the baseline values. Statistical analysis involved the non-parametric Friedman test to identify significant differences across the three time points, followed by a Wilcoxon signed-rank test with Bonferroni correction for post-hoc pairwise comparisons.

Datasets

Not discussed in the paper. The study collected data from 18 participants during a controlled experimental protocol.

Key Insights and Ideas

  • Fatigue is detectable through subtle gait changes: The study highlights that even a single IMU on the heel can capture significant kinematic changes indicative of fatigue, suggesting that complex multi-sensor systems might not always be necessary for this specific application.
  • Medial-lateral axis is a key indicator: The medial-lateral direction of movement and rotation appears to be particularly sensitive to fatigue. Changes in acceleration, angular velocity, and angle in this axis showed significant differences over the walking duration.
  • Variability and energy consumption are crucial: Beyond simple measures like range of motion, increased variability and energy consumption in gait parameters are strong indicators of fatigue and potential instability. This suggests that the body expends more effort to maintain balance as fatigue sets in.
  • Real-time monitoring potential: By collecting data continuously throughout the walking task, the study bypasses the limitations of studies that measure fatigue states in isolation, offering a more realistic approach to developing real-time fatigue detection systems.

Key Results

  • Subjective Fatigue: 16 out of 18 participants reported increased fatigue after the 60-minute walk.
  • Significant Kinematic Changes: The Friedman test revealed statistically significant differences in 25 outcome variables across the three time points.
  • Medial-Lateral Axis Dominance: Significant differences were observed in:
  • Median acceleration (p = 0.001)
    • Variability of angular velocity (p = 0.025)
    • Range of angle rotation (p = 0.0011)
  • Other Significant Findings:
    • Heel pronation angle (p = 0.005) showed significant changes.
  • Variability and energy consumption of angles in the anterior-posterior (p = 0.028), medial-lateral (p = 0.014), and vertical (p = 0.002) axes also showed significant differences.
  • Specific Parameter Trends: For example, the RMS and EC of angular velocity in the x-axis (medial-lateral) increased after 60 minutes, and the median of acceleration in the x-axis also increased.

Limitations and Future Work

The authors acknowledge several limitations:

  • Participant Selection: The study participants were generally physically active and fit older adults, which might limit the generalizability of the findings to less active or more frail elderly populations. The physical fitness of the participants could be a confounding factor.
  • Definition of Fatigue: The study relied on subjective feedback to define fatigue. In practice, fatigue is a complex phenomenon encompassing mental, neuromuscular, and metabolic aspects, and it was difficult to definitively evaluate and categorize. The study assumed an indirect relationship between fatigue and instability.
  • Single IMU Placement: While a strength for simplicity, a single IMU might not capture all aspects of gait.

Future work suggested includes:

  • Broader Participant Demographics: Sampling elderly individuals with varying levels of physical fitness and fall risk would improve external validity.
  • Objective Fatigue Measures: Incorporating objective measures like electromyography (EMG) or a musculoskeletal model could provide a more robust assessment of fatigue.
  • Algorithm Validation: Developing and validating algorithms for precise fatigue prediction, potentially using receiver-operating characteristic (ROC) curves.
  • Multi-Sensor Integration: Exploring the integration of multi-unit synchronized systems for more comprehensive activity monitoring.
  • Biofeedback Systems: Developing biofeedback systems to alert elderly individuals to their fatigue or instability, helping them minimize falling risks.

Summary updated

BibTeX
@misc{zhang2020identifying,
  author = {Guoxin Zhang and Ivy Kwan-Kei Wong and Tony Lin-Wei Chen and Tommy Tung-Ho Hong and Duo Wai-Chi Wong and Yinghu Peng and Fei Yan and Yan Wang and Qitao Tan and Ming Zhang},
  title = {{Identifying Fatigue Indicators Using Gait Variability Measures: A Longitudinal Study on Elderly Brisk Walking}},
  year = {2020},
  doi = {10.3390/s20236983},
}
Figure