Intensity normalization of two-photon microscopy images for liver fibrosis analysis

Abstract

This paper presents an intensity normalization method for analysis of liver tissue images, acquired using the two-photon microscopy system at different stages of fibrosis. Image informatics methods require precise intensity segmentation for analysis of collagen, vessel and cellular structures. Intensities of the images recorded at different time intervals corresponding to the progression of fibrosis could vary spatially and temporally depending on the experimental conditions. These variations significantly affect the image segmentation process and thus the final image analysis, especially when automatic computer-based methods are used for diagnostic parameters quantification. We propose an adaptive intensity normalization method that facilitates spatial and temporal intensity variations of the images before the segmentation process. The images are first portioned into a tessellation of regions with relatively uniform background pixels intensities and then the normalization is performed to make sure the intensity range is unified throughout the whole set of image data. This approach is further extended for montage of images acquired from multianode photomultiplier tube based multifocal multiphoton microscope (MMM) system. The proposed approach significantly improves the automated analysis of images with varying intensities without any user intervention.