A Novel Focal Tversky Loss Function with Improved Attention U-Net for Lesion Segmentation

We propose the Focal Tversky Loss, a generalized loss function that addresses extreme class imbalance in medical image segmentation. Combined with an improved Attention U-Net architecture, the method achieves state-of-the-art results on small lesion segmentation tasks and has since been widely adopted in the medical imaging community.

In IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019), pp. 683–687