Dense Dilated Multi-Scale Supervised Attention-Guided Network for histopathology image segmentation

Rangan Das, Shirsha Bose, Ritesh Sur Chowdhury, Ujjwal Maulik

Abstract: Highlights•We propose a novel attention mechanism to extract the spatial and spectral features.•A pyramid-attention module is proposed that extracts contextually relevant features.•Using deep supervision at every encoding and decoding step improves model accuracy.•Boundary segmentation mechanism helps with challenging nuclei or gland segmentation.•Dilated convolutions provide significant uplift to segmentation performance.

External IDs:dblp:journals/cbm/DasBCM23