2.5. Statistical Analysis

Statistical analyzes were performed in R software (v4.0.3, R Foundation for Statistical Computing) by K.L.R. For the calculation of the linear mixed model (LMM) as detailed below, the statistical software SPSS (v27, SPSS Inc., Chicago, IL, USA) was used.

Descriptive statistics for the SL and LT gap widths, wrist angles, and DSCs were calculated for all volunteers. Unless otherwise specified, data are given as means ± standard deviations (SDs).

To evaluate segmentation performance, the following comparisons were made: framework vs. manual segmentation 1 (reader 1), framework vs. manual segmentation 2 (reader 2), and manual segmentation 1 vs. manual segmentation 2.

Bland–Altman plots were used to visualize and comparatively evaluate automatically and manually determined SL and LT gap widths as well as wrist angles.

An LMM was used to evaluate the SL and LT gap widths based on multivariable statistics. The model included a subject-specific factor, the factors gender and wrist side, and the covariates age and height. The measurement repetition of each hand was calculated using the first-order autoregressive moving average model [42]. The model was fitted using the constrained maximum likelihood approach [43,44].

Due to the exploratory design of this study and the large amount of statistically relevant data, the significance level was set to p ≤ 0.01. In addition, alpha error accumulation was countered using the strict Bonferroni correction. This “lower-than-usual” significance level was chosen to prevent inflation of the alpha error while maintaining statistical power and reducing the false-negative rate.