A Multi-Organ Nucleus Segmentation Challenge

Abstract

Generalized nucleus segmentation techniques can contribute greatly to reducing the time to develop and validate visual biomarkers for new digital pathology datasets. We summarize the results of MoNuSeg 2018 Challenge whose objective was to develop generalizable nuclei segmentation techniques in digital pathology. The challenge was an official satellite event of the MICCAI 2018 conference in which 32 teams with more than 80 participants from geographically diverse institutes participated. Contestants were given a training set with 30 images from seven organs with annotations of 21,623 individual nuclei. A test dataset with 14 images taken from seven organs, including two organs that did not appear in the training set was released without annotations. Entries were evaluated based on average aggregated Jaccard index (AJI) on the test set to prioritize accurate instance segmentation as opposed to mere semantic segmentation. More than half the teams that completed the challenge outperformed a previous baseline [1]. Among the trends observed that contributed to increased accuracy were the use of color normalization as well as heavy data augmentation. Additionally, fully convolutional networks inspired by variants of U-Net [2], FCN [3], and Mask- RCNN [4] were popularly used, typically based on ResNet [5] or VGG [6] base architectures. Watershed segmentation on predicted semantic segmentation maps was a popular post-processing strategy. Several of the top techniques compared favorably to an individual human annotator and can be used with confidence for nuclear morphometrics.

Prognostic value of immune score in nasopharyngeal carcinoma using digital pathology

“…Then, stain deconvolution was performed, each nucleus in the haematoxylin channel was segmented using a regional convolutional neural network, which showed good performance in our previous studies ( online supplementary figure S1 ). 16 17 All the nuclei in the slide were classified into tumor cells (TCs) or tumor-associated immune cells (TAICs) based on a deep learning method, which was developed based on the manual annotation of each nucleus as a TC or TAIC by two pathologists and achieved good performance. 18 The computational algorithms showed a high reproducibility and consistency with pathological classification.…”

Cellpose: a generalist algorithm for cellular segmentation

“…We used image set MoNuSeg, which consists of 30 H&E stained histology images of various human organs [43]. Because the images contain many small nuclei (∼700 per image), we divided each of these images into 9 separate images.…”

InstantDL - An easy-to-use deep learning pipeline for image segmentation and classification

“…The second task was the multi-organ nuclei segmentation challenge. Here, 30 microscopy images of various organs with hematoxylin and eosin staining are provided (Kumar et al 2020(Kumar et al , 2017 . We reached a median Jaccard score of 0.57 (25-75%ile: 0.56 to 0.59) with InstantDL's semantic segmentation and 0.29 (25-75%ile: 0.28 to 0.30) with instance segmentation.…”

“…Boxes indicate the median and the 25/75%ile of the distribution, whiskers indicate the 1.5 interquartile range. (B) For the challenge of segmenting nuclei in microscopy images of multiple organs with hematoxylin and eosin staining (Kumar et al 2020(Kumar et al , 2017 , the winner achieved a Jaccard index of 0.69 (solid line) and the median participant 0.63 (dotted line). InstantDL using instance segmentation reached a Jaccard index of 0.29, and 0.57 using semantic segmentation.…”