Table 4.

AUC (Mean ± STD) on the outer folds of the models trained for 3 different segmentations (1 automatic and 2 manual) and 7 metrics sets.

	Exp 1: clinical	Exp 2: clinical + imaging scores	Exp 3: volume total lesion	Exp 4: volume per type	Exp 5: volume per spatial location	Exp 6: concatenation volume per type and volume per spatial location	Exp 7: volume per type and spatial location
4 classes BLAST-CT segmentation	54 ± 24	66 ± 24	59 ± 24	61 ± − 27	65 ± 25	64 ± 25	60 ± 23
4 classes manual segmentation			63 ± 22	69 ± 24	71 ± 25	68 ± 26	74 ± 26
7 classes manual segmentation				75 ± 27		73 ± 24	89 ± 17

Results from the nested cross-validation procedure on data from the Train and Validation sub-datasets of Dataset1 (23 patients). Best result in bold.