TABLE 2.

Results of therapy outcome prediction in terms of accuracy, sensitivity, specificity, and AUC for different models

	Validation set				Independent test set
Network	Acc. (%)	Sens. (%)	Spec. (%)	AUC	Acc. (%)	Sens. (%)	Spec. (%)	AUC
MLP/Clinical features	60	50	66.7	0.65	67.5	65	70	0.68
InceptionResNet	66.7	66.7	66.7	0.69	70	65	74	0.72
InceptionResNet + Seq2Seq	73.3	66.7	77.8	0.76	77.5	76.5	78.2	0.81
InceptionResNet + Transformer	73.3	66.7	77.8	0.75	77.5	76.5	78.2	0.81
InceptionResNet + LSTM	80	83.3	77.8	0.83	80	76.5	82.6	0.83
InceptionResNet + Seq2Seq with Clinical Feature Fusion	80	83.3	77.8	0.81	77.5	70.6	82.6	0.8
InceptionResNet + Transformer with Clinical Feature fusion	80	83.3	77.8	0.82	77.5	70.6	82.6	0.82
InceptionResNet + LSTM with Clinical Feature Fusion	86.7	83.3	88.9	0.88	82.5	76.5	87	0.86