Inter- and intra-reader reproducibility of shear wave elastography measurements for musculoskeletal soft tissue masses

To determine inter- and intra-reader reproducibility of shear wave elastography measurements for musculoskeletal soft tissue masses. In all, 64 patients with musculoskeletal soft tissue masses were scanned by two readers prior to biopsy; each taking five measurements of shear wave velocity (m/s) and stiffness (kPa). A single lesion per patient was scanned in transverse and cranio-caudal planes. Depth measurements (cm) and volume (cm${}^3$) were recorded for each lesion, for each reader. Linear mixed modelling was performed to assess limits of agreement (LOA), inter- and intra-reader repeatability, including analyses for measured depth and volume. Of the 64 lesions scanned, 24 (38%) were malignant. Bland-Altman plots demonstrated negligible bias with wide LOA for all measurements. Transverse velocity was the most reliable measure - intraclass correlation (95% CI) = 0.917 (0.886, 1) - though reader 1 measures could be between 38% lower and 57% higher than reader 2 [ratio-scale bias (95% LOA) = 0.99 (0.64, 1.55)]. Repeatability coefficients indicated most disagreement resulted from poor within-reader reproducibility. LOA between readers calculated from means of five repeated measurements were narrower - transverse velocity ratio-scale bias (95% LOA) = 1.00 (0.74, 1.35). Depth affected both estimated velocity and repeatability; volume also affected repeatability. This study found poor repeatability of measurements with wide LOA due mostly to intra-reader variability. Transverse velocity was the most reliable measure; variability may be affected by lesion depth. At least five measurements should be reported with LOA to assist future comparability between shear wave elastography systems in evaluating soft tissue masses.