To carry out the ablation study, we selected two key components, i.e. the coarse segmentation module and the domain transfer module, to analyze their contribution to lesion malignancy classification using T2 images. We compared our CMD²A-Net with its two variants using AUC, i.e. (1) CMD²A-Net excluding domain transfer module (i.e. CM-Net, shown in the black dashed box in Figure 4 ) and (2) CMD²A-Netexcluding the course segmentation modules (D²A-Net). Since the CM-Net does not contain DA modules, it was trained in the source domain, then fine-tuned and tested in the target domain. Datasets, P-x and LC-A, were selected as the source and target domain, respectively. D²A-Net obtains a lower AUC (0.65) compared with CMD²A-Net (0.87). This suggests that the coarse segmentation module is essential for domain-invariant feature extraction between domains. This also supports our hypothesis that the coarse lesion maps would enhance the malignancy classification accuracy. CM-Net obtains an AUC of 0.67, also less than CMD²A-Net. This indicates that the domain transfer module can substantially mitigate domain shift, thus enhancing CMD²A-Net’s PCa classification performance.
The loss parameters sensitivity is also analyzed. CMD²A-Net was trained using P-x (source domain) and LC-A (target domain). Hyperparameters \(\alpha\) and \(\beta\) (i.e. weighting parameters of the total loss) would influence the model’s generalization ability essentially. Figure 3d shows the AUC of our model with various values of  \(\alpha\) and \(\beta\) , both of which ranged in {0, 0.1, 0.05, 0.1, 0.5, 1.0, 5.0}. We can observe that CMD²A-Net demonstrates superior classification performance with   \(\alpha\) within [0.1, 1.0] and  \(\beta\) within [0.05, 1.0]. It should be noted that our model receives the lowest AUC when either  \(\alpha\) or \(\beta\) is set to 0, showing that the coarse segmentation module and the domain transfer module could enhance cross-domain knowledge transferability positively, thus improving lesion classification accuracy.