NuInsSeg: A fully annotated dataset for nuclei instance segmentation in H&E-stained histological images

Area of Science:

• Digital pathology
• Histopathology image analysis
• Computational biology

Background:

• Automatic nuclei instance segmentation is crucial for whole slide image analysis in computational pathology.
• Supervised deep learning (DL) methods excel in segmentation but require extensive annotated data, which is difficult to obtain in the medical field.

Purpose of the Study:

• To introduce NuInsSeg, a comprehensive dataset for nuclei instance segmentation.
• To provide a valuable resource for training and evaluating DL models in histopathology.
• To address the challenge of data annotation by including ambiguous area masks.

Main Methods:

• Creation of a large-scale dataset (NuInsSeg) with manually annotated nuclei.
• Inclusion of ambiguous area masks to represent regions of uncertain annotation.
• Dataset comprises 665 image patches from 31 human and mouse organs, featuring over 30,000 segmented nuclei.

Main Results:

• Development of one of the largest manually annotated nuclei datasets for H&E-stained images.
• Introduction of ambiguous area masks, a novel annotation type for challenging regions.
• Public release of the dataset and annotation instructions to facilitate research.

Conclusions:

• NuInsSeg serves as a significant resource for advancing nuclei instance segmentation in computational pathology.
• The inclusion of ambiguous areas enhances the dataset's utility for robust model development.
• Public availability promotes reproducibility and further research in histopathology image analysis.