Milken Institute School of Public Health Poster Presentations (Marvin Center & Video)

Title

Statistical modeling of musculoskeletal ultrasound images reveals correlates of age-related muscle performance

Poster Number

68

Document Type

Poster

Status

Staff

Abstract Category

Exercise and Nutrition Sciences

Keywords

aging, sarcopenia, ultrasound, screening, muscle

Publication Date

4-2017

Abstract

Background: Musculoskeletal ultrasound (MUS) is an inexpensive method to assess age-related changes in muscle tissue composition. Statistical modeling of MUS image data is relatively unexplored and may reveal correlates of muscle function and quality. The primary objective of this study was to determine how well several statistical models fit MUS image data. The secondary objective was to assess the association between model parameters and muscle performance in young and older adults.

Methods: Seventeen young (age = 24 yrs. ± 2) and seventeen older (age = 65 yrs. ± 7) adults enrolled in the study. Ultrasound scans of the rectus femoris muscle were obtained using B-mode MUS with a 13-6 MHz linear array transducer. For each scan, grayscale data were extracted from a region encompassing the muscle, and parameters for the normal, Poisson, and negative binomial distributions were estimated. Theoretical data were generated from parameter estimates, and R2 values were computed to assess agreement between grayscale and theoretical data. A one-way ANOVA was used to test for differences between each statistical model and F-tests were performed to compare goodness-of-fit. Muscle performance was measured with a hand dynamometer, and correlation analysis was conducted to determine the association between hand grip strength and parameter estimates.

Results: Mean R2 values were similar between the negative binomial (R2 = 0.93 ± 0.06) and normal (R2 = 0.84 ± 0.10) distributions (p = 0.141) and both demonstrated good agreement with grayscale data. The Poisson distribution had poor agreement (R2 = -0.34 ± 0.60) was dissimilar to the other models (p < 0.001), and was excluded from further analysis. Fit between grayscale and theoretical data was statistically better using the negative binomial distribution compared to the normal distribution for all ultrasound scans (mean F253,253: 2.70 ± 1.10, p: < 0.0001). Hand grip strength was strongly associated with negative binomial dispersion parameter estimates in older (R2 = 0.80), but not young (R2 = 0.17), adults. Mean grayscale values were moderately associated with hand grip strength in both young (R2 = 0.39) and older (R2 = 0.39) adults.

Conclusions: MUS data are best modeled by the negative binomial distribution, and dispersion parameter estimates could be used to assess loss of muscle quality with age. Future work is needed to determine whether dispersion parameter estimates are associated with measures of muscle quality attained using other imaging modalities and to explore if our findings generalize to other muscle groups.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Open Access

1

Comments

To be presented at GW Annual Research Days 2017.

This document is currently not available here.

Share

COinS
 

Statistical modeling of musculoskeletal ultrasound images reveals correlates of age-related muscle performance

Background: Musculoskeletal ultrasound (MUS) is an inexpensive method to assess age-related changes in muscle tissue composition. Statistical modeling of MUS image data is relatively unexplored and may reveal correlates of muscle function and quality. The primary objective of this study was to determine how well several statistical models fit MUS image data. The secondary objective was to assess the association between model parameters and muscle performance in young and older adults.

Methods: Seventeen young (age = 24 yrs. ± 2) and seventeen older (age = 65 yrs. ± 7) adults enrolled in the study. Ultrasound scans of the rectus femoris muscle were obtained using B-mode MUS with a 13-6 MHz linear array transducer. For each scan, grayscale data were extracted from a region encompassing the muscle, and parameters for the normal, Poisson, and negative binomial distributions were estimated. Theoretical data were generated from parameter estimates, and R2 values were computed to assess agreement between grayscale and theoretical data. A one-way ANOVA was used to test for differences between each statistical model and F-tests were performed to compare goodness-of-fit. Muscle performance was measured with a hand dynamometer, and correlation analysis was conducted to determine the association between hand grip strength and parameter estimates.

Results: Mean R2 values were similar between the negative binomial (R2 = 0.93 ± 0.06) and normal (R2 = 0.84 ± 0.10) distributions (p = 0.141) and both demonstrated good agreement with grayscale data. The Poisson distribution had poor agreement (R2 = -0.34 ± 0.60) was dissimilar to the other models (p < 0.001), and was excluded from further analysis. Fit between grayscale and theoretical data was statistically better using the negative binomial distribution compared to the normal distribution for all ultrasound scans (mean F253,253: 2.70 ± 1.10, p: < 0.0001). Hand grip strength was strongly associated with negative binomial dispersion parameter estimates in older (R2 = 0.80), but not young (R2 = 0.17), adults. Mean grayscale values were moderately associated with hand grip strength in both young (R2 = 0.39) and older (R2 = 0.39) adults.

Conclusions: MUS data are best modeled by the negative binomial distribution, and dispersion parameter estimates could be used to assess loss of muscle quality with age. Future work is needed to determine whether dispersion parameter estimates are associated with measures of muscle quality attained using other imaging modalities and to explore if our findings generalize to other muscle groups.