The reliability of the measurement of muscle volume using magnetic resonance imaging in typically developing infants by two raters

Abstract

To assess intra-rater and inter-rater reliability of the manual segmentation of Magnetic Resonance Imaging (MRI) for the in vivo measurement of infant muscle volume of the knee extensor and flexor muscles by two raters. Muscles of the knee extensor and flexor muscle of ten typically developing infants (86 days ± 7 days) were scanned with MRI (Proton density sequence). Scans were then segmented using Slicer software, and volumes rendered by two raters. Intra-rater and inter-rater reliability were assessed using intra-class correlation (ICC), with mean difference (MD), standard error of the mean (SEM), and minimal detectable change (MDC) for each muscle calculated. ICCs for Intra-rater reliability of the segmentation process for the muscle volume of the muscles of the knee extensors and flexor muscles were 0.901-0.972, and 0.776-0.945 respectively, with inter-rater reliabilities between 0.914-0.954 and 0.848-0.978, for the knee extensor and flexors muscles respectively. For intra-rater reliability, MD ≤ - 0.47 cm³, MDCs for were < 1.09 cm³ and for inter-rater MD ≤ - 1.40 cm³, MDCs for were < 1.63 cm³ for all muscles. MRI segmentation for muscle volumes showed good to excellent reliability, though given the small volumes of the muscles themselves, variations between raters are amplified. Care should be taken in the reporting and interpretation of infant muscle volume.

Figure 1

(A) Cross-sectional view of the muscles of an infant thigh, the outline of individual muscles are traced along its length to create a 3D model of the muscles. (B) 3D model from the anterior view. (C) Coronial view of the segmented muscles. (D) Superior view showing the path of the femur bone in the 3D model.

Figure 2

A box and whisker plot showing the percentage inter-observer variation for each individual muscle, with statistically significant difference (p < 0.05) between the group median percentages.

Figure 3

A coronal view of a left thigh (the posterior view) highlighting the challenges faced in segmenting muscles sharing intimate borders, here the proximal region of long head of the biceps femoris and the semitendinosus muscles.