Leveraging Large-Scale Weakly Labeled Data for Semi-Supervised Mass Detection in Mammograms

Examples of mammography reports All the mammography reports used in this study are in simplified Chinese. In Figure 1 -3, we provide some examples of de-identified text reports and their English translation (automatically translated by Google Translate), together with the probabilistic labels predicted by our NLP model, namely, Mammo-RoBERTa.