While the performances are yet to achieve reliable medical diagnosis performances, we observe a positive trend with our proposed method, showing the potential of safely classifying almost 30% of samples as benign.
← all excerpts
Leveraging Multi-Task Learning to Cope With Poor and Missing Labels of Mammograms.
1
—
—