Figure 2

Evaluation of muscle mass estimates for the Nile crocodile. Comparison between individual muscles’ measurements and 3D reconstructions (a), ordered according to relative muscle mass (M_muscle/M_body = muscle mass normalised by body mass). Boxplots of the physical (dissection-based) measurements in blue, muscle volume estimates from the polygonal modelling in red and calculation from diceCT segmentation in green. Note only 13 out of 20 muscles could be unequivocally segmented, hence some are missing; see Table 2 and Supplementary Information. Boxplots of the different 3D reconstructions and physical measurements (b). The individual muscle masses were normalised by the median of the physical measurements for each respective muscle. This shows that physical measurements of mass or volume tended to be greater than polygonal modelling and diceCT segmentation, which themselves did not significantly differ. Comparison between physical measurements and polygonal modelling (c) and Bland–Altman plot of the same comparison (f); bias -0.000318 ± 0.000133. Line of equality in (c) is in dashed black and linear regression model in red (y = 0.385365 x + 0. 0.000119; adjusted R² = 0.6977). Comparison between physical measurements and diceCT segmentation (d) and Bland–Altman plot of the same comparison (g); bias -0.000335 ± 0.000187. Line of equality in (d) is in dashed black and linear regression model in red (y = 0.323222 x + 0.000096; adjusted R² = 0.6967). Comparison between diceCT segmentation and polygonal modelling (e) and Bland–Altman plot of the same comparison (h); bias 0.000062 ± 0.000049). Line of equality in (e) is in dashed black and linear regression model in red (y = 0.723366 x + 0.000145; adjusted R² = 0.7496).