A survey on automated cell tracking: challenges and solutions
Reza Yazdi,Hassan Khotanlou
DOI: https://doi.org/10.1007/s11042-024-18697-9
IF: 2.577
2024-03-09
Multimedia Tools and Applications
Abstract:Cell tracking in microscopy images is fundamental to new biological and medical discoveries today. It facilitates the study of the properties of living cells over time. Due to the temporal nature of the cell tracking task, data association is the most difficult aspect of automated cell tracking. Due to the intricate nature of biological, imaging, and algorithmic factors that influence cell segmentation and tracking results, it is challenging to provide a simple and efficient method to determine the appropriate approach for a specific dataset. For example, mitosis, a crucial biological process, plays a key role in correcting trajectories. This research looked at all the challenges of the cell tracking task and the current solutions that have been proposed so far. In this paper, we carefully identified the sources of the tracking challenges and categorized them in a hierarchical diagram to explain the impact of challenges at different levels on the different cell tracking subtasks. Then, after identifying the solutions provided so far, we classified them into three levels: strategic, tactical, and technical. At the strategy level, tracking before detection and tracking by detection are two main approaches. The tactics can be based on cell distance, similarity, overlap, motions, probability, model evaluation, and deep-learning methods. The techniques identified in our analysis include contour evolution, nearest-neighbor linking, morphological-operator-based tracking, similarity-based label propagation, overlap-based label propagation, motion-prediction-based label propagation, graph-based multiple hypothesis tracking, probability-graph-based global optimization, probability-model-based global optimization, and recently developed deep-learning models. By merging cell tracking methods at different levels in one diagram, this classification will help to understand current solutions and provide new insights for cell tracking algorithms. Overall, in this study, we conducted a comprehensive investigation of the challenges of cell tracking and its corresponding solutions, offering a unique source of information.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering