Evaluating Bone Reconstruction Accuracy

Reconstruction accuracy of pelvis and femur geometric models were evaluated using two metrics. First, the distance error between (in)complete bone segmentation (i.e. a point cloud) and the closest nodes on reconstructed bone models was calculated using the root mean square error (RMSE, mm). The RMSE is an average error which has been applied in previous studies for comparing bone reconstruction accuracy [9 (link), 16 (link), 24 (link)]. Second, the overlapping volume similarity of complete bone segmentation and reconstructed geometric bone model was calculated using the Jaccard index (%) [25 (link)]. The Jaccard index was calculated in addition to the RMSE, because it accounts for absolute differences in overlapping volume of geometric bone models. The accuracy metrics of the different bone reconstruction methods were statistically compared (p<0.05) using a repeated measure analysis of variance followed by multiple pairwise comparisons with Bonferroni adjustment in the Statistical Package for the Social Sciences version 25 (SPSS Inc., Chicago, United States).