Fig. 8.

The Dice similarity coefficients from thresholding, RF and RF + GAC from the cross-center analysis in baseline (A) and FU (B) scans (* indicating statistically significant difference). In baseline, the means and standard deviations were 0.673 ± 0.096, 0.730 ± 0.061, 0.758 ± 0.051 for thresholding, RF and RF + GAC, respectively. In FU, these values were 0.621 ± 0.102, 0.696 ± 0.072, 0.731 ± 0.061 for these three methods, respectively. RF performed significantly better than thresholding (p = 0.038 in baseline and p = 0.003 in FU) and RF + GAC performed significantly better than RF (p = 0.027 in baseline and p = 0.008 in FU).