Computation of cell paths using adaptive tracking integrated with learning based segmentation and identification of movement patterns

Abstract

The quantification and understanding of cell motility patterns are essential to elucidate the underlying regulatory pathways. The novel approaches are needed for detailed characterization of movement patterns and to build accurate motility models describing different migratory dynamics. Cell motility plays a crucial role in organogenesis, immune responses, and tissue regeneration. Many internal and external factors can affect the pathways that control cell motility. The movement is composed of several key steps, including adhesion to the substrate, contraction, protrusion events and detachment that are characterized by advanced imaging methods and can significantly vary depending on different cell types and internal conditions. Advanced microscopy techniques have enabled high-throughput acquisition of images to explore these dynamics in detail. The state-of-the-art microscopy methods have been applied for high-throughput acquisition of cell images and videos. Yet, despite technological progress, the absence of reliable automated analysis tools remains a bottleneck for characterizing cell movement. The demand for dynamic programs imposes novel approaches for computing cell tracks and their motility features. Although open-source solutions are available, most workflows involve multi-step procedures and manual tuning of parameters, which limit reproducibility and efficiency. Additionally, factors such as signal-to-noise ratio (SNR) and cell density introduce variability in imaging data, affecting the accuracy of tracking. To overcome these limitations, we implemented an integrated strategy combining learning-based segmentation with an adaptive tracking algorithm tailored for fluorescently labeled cells. This combined approach resulted in improved segmentation efficiency compared to the manual methods based on thresholding/pixel linkage. Moreover, the accuracy of cell association was enhanced when compared to the initial version of the adaptive tracking method. Furthermore, the motility parameters were computed and used to characterize the motility modes of single cells. Briefly, the persistence distribution was computed by using the cell paths obtained by adaptive tracking process. HeLa cells exhibited a random migratory pattern, whereas RPE cells displayed directional movement. In conclusion, the integration of adaptive cell tracking with watershed-based segmentation allowed for the computation of cell tracks and motility parameters.