Efficient fluorescence image normalization for time lapse movies
Autor(en):
Schwarzfischer, M.; Marr, C.; Krumsiek, J.; Hoppe, P. S.; Schroeder, T.; Theis, F. J.
Abstract:
In the last few years, single-cell time-lapse fluorescence microscopy has emerged as a key technology in the toolbox of experimental life science. Imaging fluorescently tagged proteins allows to combine future information of cellular progeny with time resolved protein dynamics. Whenever quantitative data on the intensity of the fluorescent signal is required, a careful image processing pipeline has to be applied to account for uneven illumination, background signal, varying illumination strength or photobleaching. Previous approaches commonly used an additional calibration step to infer such image characteristics by imaging fluorescent dilutions like fluorescein. Here, we describe a method to infer a time-dependent background signal and the image gain without the use of additional fluorescent substances – instead, we use the information contained in the bleaching background of the fluorescence time-lapse movie itself. First, we tile the full image into small sub-images and determine background tiles by clustering the statistical moments of the individual intensity distributions. For each image, we interpolate the full background from the identified tiles and thus reconstitute the time-dependent background image. Second, we estimate the time-independent image gain from the background tiles of all pixels and all timepoints. We are thus able to correct for a bleaching background and an uneven illumination of the experimental setup. We show the applicability of our method by comparing the intensities of fluorescent beads derived from timelapse microscopy with intensities inferred from FACS analysis. In summary, our normalization method accurately corrects for fluorescence image issues and decreases the necessary experimental work.