Surface Electromyography-Driven Parameters for Representing Muscle Mass and Strength

Joo Hye Sung, Seol-Hee Baek, Jin-Woo Park, Jeong Hwa Rho, Byung-Jo Kim2023 · DOI

TL;DR

This paper explores the use of surface electromyography (sEMG) signals to estimate muscle mass, a crucial indicator of health, especially in aging populations. By analyzing sEMG data from common isometric exercises and correlating it with established body composition measurements (BIA and ultrasound), the researchers developed a predictive model for appendicular skeletal muscle mass (ASM) that achieves high accuracy.

What Problem Does This Paper Address?

As global populations age, maintaining skeletal muscle mass and strength becomes increasingly important for overall health and independence. Sarcopenia, the age-related loss of muscle mass and strength, is a significant concern, and its diagnosis often relies on methods like DEXA, MRI, or BIA. However, these methods can be expensive, require specialized equipment, and are not always readily available. The paper addresses the need for a simpler, more accessible, and cost-effective tool for assessing muscle mass. Surface electromyography (sEMG), a non-invasive technique that measures electrical activity in muscles, has shown promise in this area, but its direct application for estimating muscle mass in a practical setting requires further investigation and validation.

Methods

The study employed a multi-modal approach to assess muscle mass and strength.

  • sEMG Data Acquisition: Surface electrodes were placed on key upper and lower limb muscles (biceps brachii, triceps brachii, rectus femoris, biceps femoris) of participants. Maximal voluntary contraction (MVC) strength was measured during isometric exercises (elbow flexion/extension, knee flexion/extension). From the sEMG signals, researchers calculated root mean square (RMS) values (MeanRMS, MaxRMS) and derived a ratio (RatioRMS) to capture different aspects of muscle electrical activity.
  • Body Composition Analysis:
    • Bioimpedance Analysis (BIA): A commercial BIA device (InBody 770) was used to measure segmental lean mass (SLM), segmental fat mass (SFM), and appendicular skeletal muscle mass (ASM).
  • Ultrasound (US): Muscle thickness of the four targeted muscles was measured using B-mode ultrasound.
  • Statistical Analysis: Pearson's correlation coefficients were used to assess the relationships between sEMG parameters, MVC strength, BIA, and US measurements. A stepwise linear regression analysis was then employed to develop a predictive model for ASM, incorporating demographic data (age, sex, height, weight) and the derived sEMG parameters. The model was validated on a separate group of participants.

Datasets

The study utilized data from 212 healthy volunteers (91 men, 121 women) aged 40-80 years. The participants were divided into a model development group (n=137) and a cross-validation group (n=75). No new benchmark datasets were introduced; rather, the study used established BIA and ultrasound measurements as ground truth for developing and validating the sEMG-based model.

Key Insights and Ideas

  • sEMG as a Proxy for Muscle Health: The study demonstrates that sEMG parameters, particularly RMS values, are significantly correlated with both muscle strength (MVC) and muscle mass indicators (BIA and ultrasound). This suggests that sEMG can serve as a practical, non-invasive proxy for assessing muscle health.
  • Muscle Strength and Mass Interplay: As expected, muscle strength was positively correlated with lean mass indicators and muscle thickness, and negatively with fat mass. This reinforces the well-established link between muscle size and force-generating capacity.
  • Predictive Power of sEMG: The developed regression model, incorporating sEMG parameters (RatioRMS(KF) and MeanRMS(EE)), along with sex, height, and weight, achieved a remarkably high R² of 0.937 for predicting ASM. This highlights the significant contribution of sEMG data to estimating muscle mass beyond traditional anthropometric measures.
  • Lower Limb sEMG Dominance in Prediction: The inclusion of lower limb sEMG parameters (specifically RatioRMS from knee flexion and MeanRMS from elbow extension) in the final ASM prediction model is noteworthy. This aligns with the understanding that sarcopenia often disproportionately affects lower limb muscles.
  • Potential for Accessible Screening: The high accuracy of the sEMG-based model suggests its potential for developing accessible and portable tools for screening and monitoring muscle mass, particularly in aging populations or in resource-limited settings.

Key Results

  • The developed ASM prediction equation is: ASM = -26.04 + 20.345 × Height + 0.178 × weight – 2.065 × (1, if female; 0, if male) + 0.327 × RatioRMS(KF) + 0.965 × MeanRMS(EE).
  • This model achieved a high adjusted R² of 0.934 and a standard error of estimate (SEE) of 1.167.
  • The model demonstrated strong predictive performance in the cross-validation group, with a correlation coefficient of r = 0.967 (p < 0.001) between measured and predicted ASM.
  • Predicted ASM values did not significantly differ from measured ASM (18.95 ± 4.33 vs. 19.18 ± 4.59; p = 0.101).
  • sEMG parameters showed significant correlations with muscle strength (MVC), with correlation coefficients ranging from 0.506 to 0.679.
  • sEMG parameters also showed significant correlations with SLM (range: 0.154 to 0.447) and muscle thickness (range: 0.136 to 0.472), and negative correlations with SFM.

Limitations and Future Work

  • Generalizability: The developed equation was based on healthy individuals without abnormal local muscle atrophy. Its applicability to populations with muscle weakness or atrophy due to neurological or musculoskeletal disorders needs further investigation.
  • Age Range: While the study included participants aged 40-80, the model's performance in very young or very old individuals might differ.
  • Number of Muscles: The study focused on four specific muscles. Evaluating a larger set of muscles could potentially improve the prediction model, though it might reduce practicality.
  • Maximal Strength Measurement: A hand-held dynamometer was used for MVC measurement, which is less standardized than laboratory-based dynamometry.
  • Statistical Model: A linear regression model was used. While common, the authors acknowledge that the interdependence of various muscle parameters might lead to an overestimation of the prediction equation's performance.

The authors suggest that further development is needed for advanced signal processing techniques and validation processes for real-world applications, acknowledging the inherent limitations of current sEMG technology. However, they conclude that sEMG holds significant potential as an assessment tool for muscle mass in middle-aged and older adults.

Summary updated

BibTeX
@misc{sung2023surface,
  author = {Joo Hye Sung and Seol-Hee Baek and Jin-Woo Park and Jeong Hwa Rho and Byung-Jo Kim},
  title = {{Surface Electromyography-Driven Parameters for Representing Muscle Mass and Strength}},
  year = {2023},
  doi = {10.3390/s23125490},
}
Figure