pcnaDeep: a fast and robust single-cell tracking method using deep-learning mediated cell cycle profiling

Abstract

Computational methods that track single-cells and quantify fluorescent biosensors in time-lapse microscopy images have revolutionised our approach in studying the molecular control of cellular decisions. One barrier that limits the adoption of single-cell analysis in biomedical research is the lack of efficient methods to robustly track single-cells over cell division events. Here, we developed an application that automatically tracks and assigns mother-daughter relationships of single-cells. By incorporating cell cycle information from a well-established fluorescent cell cycle reporter, we associate mitosis relationships enabling high fidelity long-term single-cell tracking. This was achieved by integrating a deep-learning based fluorescent PCNA signal instance segmentation module with a cell tracking and cell cycle resolving pipeline. The application offers a user-friendly interface and extensible APIs for customized cell cycle analysis and manual correction for various imaging configurations. Availability pcnaDeep is an open-source Python application under the Apache 2.0 licence. The source code, documentation and tutorials are available at https://github.com/chan-labsite/PCNAdeep. Supplementary information Supplementary data are available at Bioinformatics online.

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

“…We also introduced a new metric called the Cell Division F1 score (CDF1), to measure the cell tracker’s ability to reliably detect cell division events and accurately assign mother-daughter cell relationships. For comparison, we benchmarked SC-Track against three other freely available cell tracking algorithms that provide similar functionalities: TrackMate 25,26 , Deepcell-tracking 27 , and pcnaDeep 28 . Initial tests focused on generating single cell tracks from nuclear masks obtained in ideal conditions, using manually corrected nuclear segmentation masks with accompanying cell cycle classifications with 5-minute temporal resolutions (Fig.…”

“…We could not perform the cell tracking experiments with the Cell Tracking Challenge dataset on pcnaDeep as it requires cell cycle class information to function 28 .…”

“…When performing multi-class instance segmentations, it is often observed that the classifications of objects occasionally switch especially when the features exhibited by objects being detected does not completely fit into a particular class or suboptimal imaging conditions lead to misclassifications of detected objects. The inherent noise in the cell classifications can pose a problem if accurate classifications of cellular states are important, such as in the quantification of cell cycle phases in an image time series 28 . To overcome this inherent problem, we developed a TrackTree Class Smoothing (TCS) algorithm that employs a probabilistic cached class smoothing approach to help accurately identify cell phase transition points.…”

From pixels to insights: Machine learning and deep learning for bioimage analysis

“…[ 60 ] DL‐based methods have shown great potential in this field, with examples of successful tracking of cell migration, division, entry, and exit of cells in the field of view, long‐term tracking of mother‐daughter relationships of single cells across cell division events, automated tracking of cancer cells and T‐cells using fluorescence‐based microscopy through the integration of a DL network with TrackMate. [ 61–63 ]…”

“…An effective tracker should have robust performance under varying conditions, such as illumination changes, camera motion, low contrast, object appearance changes, and clutter [60]. DL-based methods have shown great potential in this field, with examples of successful tracking of cell migration, division, entry, and exit of cells in the field of view, longterm tracking of mother-daughter relationships of single cells across cell division events, automated tracking of cancer cells and T-cells using fluorescence-based microscopy through the integration of a DL network with TrackMate [61][62][63]. Object classification ML and DL-based image classification have revolutionized the identification and categorization of specific objects within microscopy images.Object classification can be done using post-processed data or by training DL models on the images themselves.…”