CELL TRACKING UNDER HIGH CONFLUENCY CONDITIONS BY CANDIDATE CELL REGION DETECTION-BASED ASSOCIATION APPROACH

摘要

Automated tracking of cell population is an important elementof research and discovery in the biology field. Inthis paper, we propose a method that tracks cells underhighly confluent conditions by using the candidate cell regiondetection-based association approach. Unlike conventionalsegmentation-based association tracking methods,the proposed method uses the tracking results fromthe previous frame to segment the cell regions at the currentframe. First, candidate cell regions are detected, and whilethere may be many false positives, there are very few falsenegatives. Next, optimized detection results are selectedfrom the candidate regions and associated with the trackingresults of the previous frame by resolving a linear programmingproblem. We quantitatively evaluated the proposedmethod using a variety of sequences. Results showed thatour method has a better tracking performance than conventionalsegmentation-based association methods.