Figure 4: Data validity.

Paired t-tests (p), Pearson’s product-moment correlation coefficient (r) and regression analysis, including boxplots and regression lines through the origin, were used to test the data validity. Strong positive correlations and measurement values close to the regression line were achieved between the compared segmentation models without significant variability or inaccuracy, which proves the validity of the ground truth data. (a) The calculated difference between the ground truth segmentations (A, B) was not significant for the assessment parameter volume (p > 0.05), which is graphically shown in the boxplots, which are similar between segmentations A and B. (b) The measured volume values were localized closely along the regression line; thus, the product-moment correlation coefficient (Pearson) of the regression model volume was close to the value one, not below 0.99 (r > 0.99). (c) The calculated difference between the ground truth segmentations (A, B) was not significant for the assessment parameter voxels (p > 0.05), which is graphically shown in the boxplots, which are similar between segmentations A and B. (d) The measured voxel values were localized closely along the regression line; thus, the product-moment correlation coefficient (Pearson) in the regression model voxel was close to the value one and was not below 0.99 (r > 0.99). Note: Some of the data in Figure 4 were previously released in part¹⁷.