Joint Left Atrial Segmentation and Scar Quantification Based on a DNN with Spatial Encoding and Shape Attention

no code implementations • 23 Jun 2020 • Lei Li, Xin Weng, Julia A. Schnabel, Xiahai Zhuang

Compared to conventional binary label based loss, the proposed SE loss can reduce noisy patches in the resulting segmentation, which is commonly seen for deep learning-based methods.

Segmentation