Application of temporal texture features to automated analysis of protein subcellular locations in time series fluorescence microscope images

Protein subcellular locations, as an important property of proteins, are commonly learned using fluorescence microscopy. Previous work by our group has shown that automated analysis of 2D and 3D static images can recognize all major subcellular patterns in fluorescence micrographs, and that automated methods can be used to distinguish patterns that are subtly different. Since many proteins are in constant movement within the cell, we extended our studies to time series images, which contain both spatial and temporal information. In this paper, we present the application of a set of temporal texture features, which do not require predefining objects for tracking, to the classification of subcellular location patterns. We demonstrate that these features successfully captured new information contained in the time domain by evaluating the accuracy of automated classification of a data set of five proteins with similar location patterns.