A robust and efficient automatic method to segment maize FASGA stained stem cross section images to accurately quantify histological profile

Abstract

Background: Grasses internodes are made of distinct tissues such as vascular bundles, epidermis, rind and pith. The histology of grasses stem was largely revisited recently taking advantage of the development of microscopy combined with the development of computer-automated image analysis workflows. However, the diversity and complexity of the histological profile complicates quantification. Accurate and automated analysis of histological images thus remains challenging. Results: Herein, we present a workflow that automatically segments maize internode cross section images into 40 distinct tissues: two tissues in the epidermis, 19 tissues in the rind, 14 tissues in the pith and 5 tissues in the bundles. This level of segmentation is achieved by combining the Hue, Saturation and Value properties of each pixel and the location of each pixel in FASGA stained cross sectiona. This workflow is likewise able to highlight significant and subtle histological genotypic variations between maize internodes. The grain of precision provided by the workflow also makes it possible to demonstrate different levels of sensitivity to digestion by enzymatic cocktails of the tissues in the pith. The precision and strength of the workflow is all the more impressive because it is preserved on cross section images of other grasses such as miscanthus or sorghum. Conclusions: The fidelity of this tool and its capacity to automatically identify variations of a large number of histological profiles among different genotypes pave the way for its use to identify genotypes of interest and to study the underlying genetic bases of variations in histological profiles in maize or other species.