6) LM relies on a small number of biologically-derived adjustable parameters to achieve high accuracy tracking.įigure 1 highlights LM algorithm description. 5) LM creates fusion lineages by tracking colony or cell merges. 4) Its execution is fast enough for real-time tracking and manages memory efficiently for large datasets. 3) LM uses the overlap information between current and past frames to identify and separate cells mistakenly segmented as a single cell when cell-cell contact occurs. These properties include mother cell roundness, mother cell size, daughter size similarity, and daughter aspect ratio similarity.
2) In addition to the overlap information, LM uses biological properties measured from the segmented images to detect mitosis. The tool takes labeled segmented masks as input and outputs a cell lineage tree and a set of new labeled masks where each cell is assigned a unique global tracking number. The user has the choice of any segmentation technique including manually drawn masks as input to LM. In fact, LM is totally segmentation-independent: connecting segmentation results to the tracker does not require any change in the pipeline or any special input. It has six equally important and unique capabilities: 1) LM operates on segmented masks, therefore the system is not dependent upon a particular segmentation method. While Lineage Mapper was mainly developed for cell biology, it has also been successfully applied on particle tracking as we demonstrate in the validation/results section. LM detects 2D dynamic single cell behaviors: migration, mitosis, cell death, cells within sheets, and cells moving with high cell-cell contact. We developed Lineage Mapper (LM) to address these challenges ( Supplementary Table 2, and Supplementary Note 1). Characteristics of an algorithm that are common to most cell biology problems 14, 15, 16, 17 include the accuracy of tracking over a range of cell contact levels (from well-separated to confluent cultures), scalability, simple communication with any segmentation method, and the minimization of non-intuitive parameters that map to the underlying mathematical models. There is a need to develop tracking tools with sufficient functionality and flexibility to render themselves broadly applicable within multiple scenarios. In most available tracking techniques, a segmentation method cannot easily be replaced with another more accurate one for the researcher’s specific application. Most common cell tracking techniques 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13 are linked to a particular segmentation method, where there is inherent feedback between the segmentation and the tracking algorithms ( Supplementary Table 1), thus making them impractical for use across a broad range of applications. The related problem of tracking individual objects in time-series image data is also similarly constrained. Segmentation, the process of outlining objects of interest in digital images, is a very challenging aspect of image analysis and is usually custom-designed for a specific cell line and imaging modality. However, obtaining useful quantitative dynamic data related to cell or colony behavior (including the identification of cell growth, mitosis, migration, proliferation, death, fusion, and differentiation) requires image analysis methods that can accurately segment and track cells in the presence of frequent cell-cell contacts.Ī typical workflow to quantify single cell dynamics begins with segmentation, followed by tracking the segmented masks, and finally extracting dynamic tracking outputs, from which all post-processing analysis can be derived. Automated microscopy has facilitated the large scale acquisition of live cell image data 1 to monitor migration, morphology, and lineage tracing of large numbers of single cells or colonies in culture.
0 kommentar(er)
