Year-end Sale % OFF 
Abstract
SUMMARYSingle-cell segmentation is typically a crucial task of image-based cellular analysis. We present nucleAIzer, a deep-learning approach aiming toward a truly general method for localizing 2D cell nuclei across a diverse range of assays and light microscopy modalities. We outperform the 739 methods submitted to the 2018 Data Science Bowl on images representing a variety of realistic conditions, some of which were not represented in the training data. The key to our approach is that during training nucleAIzer automatically adapts its nucleus-style model to unseen and unlabeled data using image style transfer to automatically generate augmented training samples. This allows the model to recognize nuclei in new and different experiments efficiently without requiring expert annotations, making deep learning for nucleus segmentation fairly simple and labor free for most biological light microscopy experiments. It can also be used online, integrated into CellProfiler and freely downloaded at www.nucleaizer.org.A record of this paper’s transparent peer review process is included in the Supplemental Information.
Highlights
Identifying nuclei is the starting point for many microscopybased cellular analyses, which are widespread in biomedical research
Approaches using deep learning dominated the competition, achieving scores that shattered what was previously possible: the best performing traditional methods we submitted ranked no higher than 1,000 out of 3,891 submissions in stage 1; even classical methods hand-tuned to five subsets of the testing data were beaten by 85 out of 739 submissions in stage 2 testing (Caicedo et al, 2019b)
The top deeplearning-based methods relied on only a handful of different architectures, namely Mask R-CNN, U-Net, and feature-pyramid networks; the factors that participants commonly believed had most influence over their method’s ranking were the amount of data, the pre-processing, and methods used to augment the data
Summary
Identifying nuclei is the starting point for many microscopybased cellular analyses, which are widespread in biomedical research. The dominant approaches for this task have been based on classic image processing algorithms (e.g., thresholding and seeded watershed; Carpenter et al, 2006), guided by shape and spatial priors (Molnar et al, 2016). These methods require expert knowledge to properly adjust the parameters, which typically must be retuned when experimental conditions change. Deep learning has revolutionized an assortment of tasks in image analysis, from image classification (Krizhevsky et al, 2017) to face recognition (Taigman et al, 2014) and scene segmentation (Badrinarayanan et al, 2017). Initial work (reviewed in Moen et al, 2019) indicates that deep learning is effective for nucleus segmentation (Falk et al, 2019; Van Valen et al, 2016; Cui et al, 2018); these methods often fail to properly separate touching nuclei well and most importantly lack robustness to unseen domains
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
