Evaluation of Deep Learning Strategies for Nucleus Segmentation in Fluorescence Images

Abstract

Identifying nuclei is often a critical first step in analyzing microscopy images of cells and classical image processing algorithms are most commonly used for this task. Recent developments in deep learning can yield superior accuracy, but typical evaluation metrics for nucleus segmentation do not satisfactorily capture error modes that are relevant in cellular images. We present an evaluation framework to measure accuracy, types of errors, and computational efficiency; and use it to compare deep learning strategies and classical approaches. We publicly release a set of 23,165 manually annotated nuclei and source code to reproduce experiments and run the proposed evaluation methodology. Our evaluation framework shows that deep learning improves accuracy and can reduce the number of biologically relevant errors by half. © 2019 The Authors. Cytometry Part A published by Wiley Periodicals, Inc. on behalf of International Society for Advancement of Cytometry.

Towards a scaled-up T cell-mediated cytotoxicity assay in 3D cell culture using microscopy

“…The two other methods employed extend this process by attempting to further identify individual cells within spheroids. The first of these uses the U-Net-based nuclei segmentation model of CellProfiler (Caicedo et al 2019) via Cytokit (Czech et al 2018) while the second utilizes a custom cytometry function based on differences of gaussian filters. This second method affords greater flexibility in parameterization to aid in tuning thresholds (albeit subjectively) to better capture nuclei images with noisy, occluded boundaries.…”

AnnotatorJ: an ImageJ plugin to ease hand-annotation of cellular compartments

“…Recently, deep learning (DL) methods have proven themselves worthy of consideration in microscopy image analysis tools as they have also been successfully applied in a wider range of applications including but not limited to face detection (Taigman et al, 2014, Sun et al 2014, Schroff et al 2015, self-driving cars (Badrinarayanan et al, 2017, Redmon et al, 2016, Grigorescu et al, 2019 and speech recognition (Hinton et al, 2012). Caicedo et al (Caicedo et al, 2019) and others (Hollandi et al 2019, Moshkov et al 2019 proved that single cell detection and segmentation accuracy can be significantly improved utilizing DL networks. The most popular and widely used deep convolutional neural networks (DCNNs) include Mask R-CNN (He et al, 2017): an object detection and instance segmentation network, YOLO (Redmon et al, 2016): a fast object detector and U-Net (Ronneberger et al, 2015): a semantic segmentation approach specifically intended for bioimage analysis purposes.…”

Practical Fluorescence Reconstruction Microscopy for Large Samples and Low-Magnification Imaging

“…The U-Net itself is 50 commonly used in machine learning approaches because it is a lightweight convolutional neural 51 network (CNN) which readily captures information at multiple spatial scales within an image, 52 thereby preserving reconstruction accuracy while reducing the required number of training 53 samples and training time. U-Nets, and related deep learning approaches, have found broad 54 application to live-cell imaging tasks such as cell phenotype classification, feature 55 segmentation 10, [14][15][16][17][18][19] , and histological stain analysis [20][21][22][23] . 56 57…”