. Author manuscript; available in PMC: 2021 Oct 26.

Published in final edited form as: Eur Radiol. 2020 Oct 1;31(4):2559–2567. doi: 10.1007/s00330-020-07274-x

Table 1.

Accuracy to classify three molecular subtypes in Training and Testing datasets using CNN and CLSTM

Dataset	Process	CNN	CLSTM
Training Dataset	Initial Training^*	0.73–0.89 (0.79)	0.83–0.95 (0.91)
Testing Dataset-1	Testing Using the First Trained Model	0.52	0.44
	Second Training Using Transfer Learning^*	0.85–0.95 (0.91)	0.79–0.88 (0.83)
	Testing Using the Second Model from Transfer Learning of Dataset-2	0.82	0.76
Testing Dataset-2	Testing Using the First Model	0.47	0.39
	Second Training Using Transfer Learning^*	0.82–0.89 (0.85)	0.74–0.87 (0.82)
	Testing Using the Second Model from Transfer Learning of Dataset-1	0.78	0.74

The accuracy in the training process is evaluated using 10-fold cross-validation, and the range (mean) is shown