Weakly Supervised Deep Nuclei Segmentation Using Partial Points Annotation in Histopathology Images.

Area of Science:

• Digital Pathology
• Computational Biology
• Medical Image Analysis

Background:

• Nuclei segmentation is crucial for histopathology image analysis.
• Manual pixel-wise annotation for supervised training is labor-intensive.
• Weakly supervised learning offers a potential solution to reduce annotation burden.

Purpose of the Study:

• To develop a novel weakly supervised nuclei segmentation framework using partial point annotations.
• To reduce the significant manual effort required for accurate annotation in histopathology.
• To achieve competitive segmentation performance with substantially less annotation.

Main Methods:

• A two-stage framework involving nuclei detection and segmentation.
• Stage 1: Semi-supervised learning with an extended Gaussian mask and self-training with background propagation for nuclei detection.
• Stage 2: Weakly supervised segmentation using derived coarse labels and a fully-connected conditional random field loss.

Main Results:

• The proposed method achieves competitive performance compared to fully supervised methods.
• It outperforms state-of-the-art methods on two nuclei segmentation datasets.
• The framework significantly reduces the required annotation effort.

Conclusions:

• The novel weakly supervised framework effectively performs nuclei segmentation with partial point annotations.
• This approach alleviates the need for extensive manual annotation in histopathology.
• The method offers a practical and efficient alternative for nuclei segmentation tasks.