An Auxiliary Task for Learning Nuclei Segmentation in 3D Microscopy Images

Peter Hirsch, Dagmar Kainmueller
Proceedings of the Third Conference on Medical Imaging with Deep Learning, PMLR 121:304-321, 2020.

Abstract

Segmentation of cell nuclei in microscopy images is a prevalent necessity in cell biology. Especially for three-dimensional datasets, manual segmentation is prohibitively time-consuming, motivating the need for automated methods. Learning-based methods trained on pixel-wise ground-truth segmentations have been shown to yield state-of-the-art results on 2d benchmark image data of nuclei, yet a respective benchmark is missing for 3d image data. In this work, we perform a comparative evaluation of nuclei segmentation algorithms on a database of manually segmented 3d light microscopy volumes. We propose a novel learning strategy that boosts segmentation accuracy by means of a simple auxiliary task, thereby robustly outperforming each of our baselines. Furthermore, we show that one of our baselines, the popular three-label model, when trained with our proposed auxiliary task, outperforms the recent {\em StarDist-3D}. As an additional, practical contribution, we benchmark nuclei segmentation against nuclei {\em detection}, i.e. the task of merely pinpointing individual nuclei without generating respective pixel-accurate segmentations. For learning nuclei detection, large 3d training datasets of manually annotated nuclei center points are available. However, the impact on detection accuracy caused by training on such sparse ground truth as opposed to dense pixel-wise ground truth has not yet been quantified. To this end, we compare nuclei detection accuracy yielded by training on dense vs. sparse ground truth. Our results suggest that training on sparse ground truth yields competitive nuclei detection rates.

Cite this Paper

BibTeX
@InProceedings{pmlr-v121-hirsch20a, title = {An Auxiliary Task for Learning Nuclei Segmentation in 3D Microscopy Images}, author = {Hirsch, Peter and Kainmueller, Dagmar}, booktitle = {Proceedings of the Third Conference on Medical Imaging with Deep Learning}, pages = {304--321}, year = {2020}, editor = {Arbel, Tal and Ben Ayed, Ismail and de Bruijne, Marleen and Descoteaux, Maxime and Lombaert, Herve and Pal, Christopher}, volume = {121}, series = {Proceedings of Machine Learning Research}, month = {06--08 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v121/hirsch20a/hirsch20a.pdf}, url = {https://proceedings.mlr.press/v121/hirsch20a.html}, abstract = {Segmentation of cell nuclei in microscopy images is a prevalent necessity in cell biology. Especially for three-dimensional datasets, manual segmentation is prohibitively time-consuming, motivating the need for automated methods. Learning-based methods trained on pixel-wise ground-truth segmentations have been shown to yield state-of-the-art results on 2d benchmark image data of nuclei, yet a respective benchmark is missing for 3d image data. In this work, we perform a comparative evaluation of nuclei segmentation algorithms on a database of manually segmented 3d light microscopy volumes. We propose a novel learning strategy that boosts segmentation accuracy by means of a simple auxiliary task, thereby robustly outperforming each of our baselines. Furthermore, we show that one of our baselines, the popular three-label model, when trained with our proposed auxiliary task, outperforms the recent {\em StarDist-3D}. As an additional, practical contribution, we benchmark nuclei segmentation against nuclei {\em detection}, i.e. the task of merely pinpointing individual nuclei without generating respective pixel-accurate segmentations. For learning nuclei detection, large 3d training datasets of manually annotated nuclei center points are available. However, the impact on detection accuracy caused by training on such sparse ground truth as opposed to dense pixel-wise ground truth has not yet been quantified. To this end, we compare nuclei detection accuracy yielded by training on dense vs. sparse ground truth. Our results suggest that training on sparse ground truth yields competitive nuclei detection rates.} }
Endnote
%0 Conference Paper %T An Auxiliary Task for Learning Nuclei Segmentation in 3D Microscopy Images %A Peter Hirsch %A Dagmar Kainmueller %B Proceedings of the Third Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2020 %E Tal Arbel %E Ismail Ben Ayed %E Marleen de Bruijne %E Maxime Descoteaux %E Herve Lombaert %E Christopher Pal %F pmlr-v121-hirsch20a %I PMLR %P 304--321 %U https://proceedings.mlr.press/v121/hirsch20a.html %V 121 %X Segmentation of cell nuclei in microscopy images is a prevalent necessity in cell biology. Especially for three-dimensional datasets, manual segmentation is prohibitively time-consuming, motivating the need for automated methods. Learning-based methods trained on pixel-wise ground-truth segmentations have been shown to yield state-of-the-art results on 2d benchmark image data of nuclei, yet a respective benchmark is missing for 3d image data. In this work, we perform a comparative evaluation of nuclei segmentation algorithms on a database of manually segmented 3d light microscopy volumes. We propose a novel learning strategy that boosts segmentation accuracy by means of a simple auxiliary task, thereby robustly outperforming each of our baselines. Furthermore, we show that one of our baselines, the popular three-label model, when trained with our proposed auxiliary task, outperforms the recent {\em StarDist-3D}. As an additional, practical contribution, we benchmark nuclei segmentation against nuclei {\em detection}, i.e. the task of merely pinpointing individual nuclei without generating respective pixel-accurate segmentations. For learning nuclei detection, large 3d training datasets of manually annotated nuclei center points are available. However, the impact on detection accuracy caused by training on such sparse ground truth as opposed to dense pixel-wise ground truth has not yet been quantified. To this end, we compare nuclei detection accuracy yielded by training on dense vs. sparse ground truth. Our results suggest that training on sparse ground truth yields competitive nuclei detection rates.
APA
Hirsch, P. & Kainmueller, D.. (2020). An Auxiliary Task for Learning Nuclei Segmentation in 3D Microscopy Images. Proceedings of the Third Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 121:304-321 Available from https://proceedings.mlr.press/v121/hirsch20a.html.

Related Material