Bending Loss Regularized Network for Nuclei Segmentation in Histopathology Images

TL;DR

This paper introduces a bending loss regularized network that improves nuclei segmentation in histopathology images by effectively handling overlapped nuclei, outperforming existing methods on multiple metrics.

Contribution

The paper proposes a novel bending loss function that penalizes high curvature contours, enhancing segmentation accuracy for overlapped nuclei in histopathology images.

Findings

Outperforms six state-of-the-art methods on MoNuSeg dataset

Achieves higher scores on Aggregate Jaccard Index and Dice

Effectively segments overlapped nuclei

Abstract

Separating overlapped nuclei is a major challenge in histopathology image analysis. Recently published approaches have achieved promising overall performance on public datasets; however, their performance in segmenting overlapped nuclei are limited. To address the issue, we propose the bending loss regularized network for nuclei segmentation. The proposed bending loss defines high penalties to contour points with large curvatures, and applies small penalties to contour points with small curvature. Minimizing the bending loss can avoid generating contours that encompass multiple nuclei. The proposed approach is validated on the MoNuSeg dataset using five quantitative metrics. It outperforms six state-of-the-art approaches on the following metrics: Aggregate Jaccard Index, Dice, Recognition Quality, and Pan-optic Quality.