Table 5.

Results of the breast cancer risk prediction methods proposed in [29]. The expression e.g., “Prior-4 → Prior-3”, denotes transfer learning of a model trained on Prior-4 for training the model on Prior-3.

Method	CC view AUC	MLO view AUC
Prior data-enabled transfer learning (Prior-4 → Prior-3)	0.54	0.55
Prior data-enabled transfer learning (Prior-4 → Prior-3 → Prior-2)	0.58	0.56
Prior data-enabled transfer learning (Prior-4 → Prior-3 → Prior-2 → Prior-1)	0.61	0.60
Multi-task learning using four priors	0.60	0.59