Shearlet Analysis of Confocal Laser-Scanning Microscopy Images to Extract Morphological Features of Neurons

Abstract

Due to the progress in laser scanning microscopy techniques computational image analysis methods increasingly have come in the focus of biology. Methods of image analysis can broadly be divided into two groups: The first group deals with segmentation and noise reduction problems, while the second group focuses on the statistical and morphological analysis of structures. Structure analysis strongly depends on the quality of the output of segmentation approaches, which aim to separate the digital image in undesirable (background) and desirable (foreground) information. The progress in microscopy techniques has led to a large amount of highly detailed images containing fine structures. Unfortunately, such images cannot be separated in an automated way without loss of considerable detail and information. In this chapter, we present an approach that is based on compressed sensing methods to separate biologically relevant information, in this case the structure of dendrites, from an image background. The approach has the advantage that it allows separating even fine structural details in large images without the common disadvantages of intensity based algorithms. We have written a freely downloadable software suite and present a detailed protocol of its use to determine morphological features of dendritic trees from fluorescence stained 3D image stacks.