Structure preserving adversarial generation of labeled training samples for single-cell segmentation

Tasnádi, Ervin Áron and Sliz-Nagy, Alex and Horváth, Péter (2023) Structure preserving adversarial generation of labeled training samples for single-cell segmentation. CELL REPORTS METHODS, 3 (9). No.-100592. ISSN 2667-2375

Preview

Text TasnadiEetal.pdf - Published Version Available under License Creative Commons Attribution. Download (6MB) | Preview

Abstract

We introduce a generative data augmentation strategy to improve the accuracy of instance segmentation of microscopy data for complex tissue structures. Our pipeline uses regular and conditional generative adver-sarial networks (GANs) for image-to-image translation to construct synthetic microscopy images along with their corresponding masks to simulate the distribution and shape of the objects and their appearance. The synthetic samples are then used for training an instance segmentation network (for example, StarDist or Cell -pose). We show on two single-cell-resolution tissue datasets that our method improves the accuracy of downstream instance segmentation tasks compared with traditional training strategies using either the raw data or basic augmentations. We also compare the quality of the object masks with those generated by a traditional cell population simulation method, finding that our synthesized masks are closer to the ground truth considering Fre ' chet inception distances.

Actions (login required)

Edit Item