Comparison of manual and semi-automated algorithm for measuring architectural features during different isometric...

Ultrasound is a cost-effective and reliable method to determine skeletal muscle architecture. However, manual analysis of fascicle length (FL) and pennation angle (PA) can be arduous and subjective among raters, particularly among novice raters. Alternatives to manual processing have been proposed that expedite the evaluation of muscle architecture and afford more consistency. While using algorithms has provided dependable results of muscle architecture, it has often focused on variables of passive range of motion and submaximal contractions. To fully understand the impact of muscle architecture using semi-automated analysis, an investigation of a broad range of contraction intensities is needed. The purpose of this study was to develop and determine the intra-rater and inter-rater reliability of a custom, semi-automated algorithm to extract measures of muscle thickness, pennation angle, and fascicle length, and second to compare the semi-automated measures to measures extracted manually from the same novice raters while accounting for differences between contraction intensities.

Fifteen resistance-trained individuals (male: n = 6, female: n = 9) completed this study. Images were collected during four contraction intensities relative to maximal voluntary isometric contractions (MVIC) (at rest, 30%, 70%, and MVIC) and analyzed by three novice raters to compare the semi-automated algorithm and manual measurement in the vastus lateralis.

Intra-rater reliability for manual measures was poor for FL (ICCs: 0-0.30), poor to good for PA (ICCs: 0.46-0.77), and moderate to good for muscle thickness (MT) (ICCs: 0.55-0.84). For the semi-automated algorithm, the intra-rater reliability was good to excellent for FL (range: 0.90-0.99), PA (range: 0.88-0.99), and MT (range: 0.996-0.999) across all contraction intensities.

The findings of this study suggest that the reliability of manual measurement is lower when novice raters perform image analyses compared to the semi-automated method. Therefore, careful consideration and training should be provided when considering manual assessment of muscle architecture values, and standardized identification methods and features in algorithm development may be a better method for reproducibility.