Automated tracking in live-cell time-lapse movies

Abstract

The development of high-throughput live cell imaging is currently limited by the capabilities of image analysis. Software is required to generate single cell time courses from large data sets of time-lapse movies and to follow properties of individual cells. Automated cell tracking faces notorious problems associated with cell division, high cell density and cell mobility. In particular, a large number of cell traces are discarded in experiments with extended observation times due to image analysis ambiguities. Here we develop an algorithm for robust tracking based on cost matrices from multiple cell parameters such as object size, position or texture. Singularities in cost indicate tracking conflicts, which can be categorized into event classes such as cell division, lysis or overlap of cells. We demonstrate that multiple parameter tracking (MPT) generates single cell fluorescence time traces more reliably than algorithms based on position tracking only. Context-sensitive automatic evaluation and event management increase the yield of continuous and correctly assigned time traces by 27%.