Estimating the Duration of Overlapping Events from Image Sequences Using Cylindrical Temporal Boolean Models

摘要

Recent advances in microscopy jointly to the development of fluorescent probes have enabled to image dynamic processes with very high spatial-temporal resolution, for instance in Cell Biology. In some applications, the segmented areas associated with different events overlap spatially and temporally forming random clumps. In order to study the shape-size features and durations of the events, it is a usual practice to analyze only isolated episodes. However, this sample is biased, because faster and smaller events tend to be isolated. We model the images as a realization of a cylindrical temporal Boolean model. We evaluate the bias introduced when ruling out non-isolated episodes. We propose an estimator of the duration distribution and perform a simulation study to assess its accuracy. The method is applied to fluorescent-tagged proteins image sequences. Results show that this procedure is effective for analyzing dynamic processes where spatial and temporal overlapping occurs.