Skip to main content

Advertisement

Table 3 Results of cross-validation study

From: MRI-based treatment plan simulation and adaptation for ion radiotherapy using a classification-based approach

Contrasts Additional features Mean absolute
   error [HU]
TSE1, UTE2 box.sd, dist.center 92.5
TSE1, UTE2 box.sd, dist.xyz 93.5
TSE1, UTE1 box.sd, dist.center 94.1
TSE1, UTE2 box.sd, dist.center, dist.xyz 95.8
TSE1, UTE2 box.mean, box.sd, dist.center 96.5
TSE1, UTE2 box.mean, box.sd, dist.xyz 96.5
TSE1, MPRAGE, UTE2 box.sd, dist.center 96.7
TSE1, TSE2, UTE2 box.sd, dist.center 97.0
TSE1, UTE1, UTE2 none 111.7
UTE1 box.mean, box.sd, dist.xyz 100.7
TSE1, MPRAGE box.mean, box.sd, dist.center 115.7
  1. Mean absolute errors of masked voxels for the 8 best combinations of contrasts and additional features out of 738 combinations in comparison to the best result without any additional feature as well as without TSE1 and UTE sequences. Results were obtained for reduced datasets of 32 images per sample.