The Cell Tracking Challenge: 10 years of objective benchmarking

Abstract

The Cell Tracking Challenge is an ongoing benchmarking initiative that has become a reference in cell segmentation and tracking algorithm development. Here, we present a significant number of improvements introduced in the challenge since our 2017 report. These include the creation of a new segmentation-only benchmark, the enrichment of the dataset repository with new datasets that increase its diversity and complexity, and the creation of a silver standard reference corpus based on the most competitive results, which will be of particular interest for data-hungry deep learning-based strategies. Furthermore, we present the up-to-date cell segmentation and tracking leaderboards, an in-depth analysis of the relationship between the performance of the state-of-the-art methods and the properties of the datasets and annotations, and two novel, insightful studies about the generalizability and the reusability of top-performing methods. These studies provide critical practical conclusions for both developers and users of traditional and machine learning-based cell segmentation and tracking algorithms.

SC-Track: a robust cell tracking algorithm for generating accurate single-cell lineages from diverse cell segmentations

“…To demonstrate that SC-Track can perform well in a diverse set of cell types and imaging conditions, we expanded our tracking benchmarks to a collection of publicly available microscopy datasets (Supplementary Table 5). We used the silver reference segmentation results from the Cell Tracking Challenge (CTC) because the CTC dataset contains a wide collection of timelapse microscopy images taken with a variety of imaging settings on various cancer cells of diverse morphologies 29 . The segmentation results from the CTC dataset are equally diverse ranging from nuclear masks to whole cell segmentations.…”

“…The segmentation results from the CTC dataset are equally diverse ranging from nuclear masks to whole cell segmentations. We used the silver reference segmentation dataset since the segmentation results were derived from the best performing CNN models in the CTC 29 . Furthermore, the silver reference segmentations were accompanied by ground truth tracking results, making these datasets an impartial real-life test to measure the generalisability of SC-Track’s cell tracking algorithm.…”

“…To test the universal functionality of SC-Track to accurately generate single cell linages from a variety of cell types and segmentation modes, we used the silver reference segmentation results from the Cell Tracking Challenge 29 . The silver reference datasets represent the uncorrected segmentation results obtained from CNNs applied on a diverse variety of mammalian cell lines of different morphological appearances and imaging conditions (Supplementary Table 5).…”

A general algorithm for consensus 3D cell segmentation from 2D segmented stacks

“…A second example is a video of MDA231 human breast carcinoma cells embedded in a collagen matrix from the single cell tracking challenge 87 (Fig. 4j).…”

Facilitating cell segmentation with the projection-enhancement network

“…The decoders predict cell distances and neighbor distances that are used for a post-processing watershed algorithm. DUNet is the state-of-the-art in many cell segmenting and tracking challenges [30]. For evaluative purposes, we used the best performing pretrained DUNet model, trained on 3D Fluo-C3DL-MDA231 dataset [31], and took MIPs of the resulting 3D detected objects.…”