Texture Feature Extraction From Microscope Images Enables Robust Estimation of ER Body Phenotype in Arabidopsis

Abstract

Abstract Background: Cellular components are controlled by genetic and physiological factors that define their shape and size. However, quantitively capturing the morphological characteristics and movement of cellular organelles from micrograph images is challenging, because the analysis deals with complexities of images that frequently lead to inaccuracy in the estimation of the features. Here we show a unique quantitative method to overcome biases and inaccuracy of biological samples from confocal micrographs. Results: We generated 2D images of cell walls and spindle-shaped cellular organelles, namely ER bodies, with a maximum contrast projection of 3D confocal fluorescent microscope images. The projected images were further processed and segmented by adaptive thresholding of the fluorescent levels in the cell walls. Micrographs are composed of pixels, which have information on position and intensity. From the pixel information we calculated three types of features (spatial, intensity and Haralick) in ER bodies corresponding to segmented cells. The spatial features include basic information on shape, e.g., surface area and perimeter. The intensity features include information on mean, standard deviation and quantile of fluorescence intensities within an ER body. Haralick features describe the texture features, which can be calculated mathematically from the interrelationship between the pixel information. Together these parameters were subjected to multivariate analysis to estimate the morphological diversity. Additionally, we calculated the displacement of the ER bodies using the positional information in a time-lapse image. We captured similar morphological diversity and movement within ER body phenotypes on several microscopy experiments performed in different settings and scanned under different objectives. We then described differences in morphology and movement of ER bodies between A. thaliana wild type and mutants deficient in ER body-related genes. Conclusions: The findings unexpectedly revealed multiple genetic factors that are involved in the shape and size of ER bodies in A. thaliana . This is the first report showing morphological characteristics in addition to the movement of cellular components and quantitatively summarises plant phenotypic differences even in plants that show similar cellular components. The estimation of morphological diversity was independent of the cell staining method and the objective lens used in the microscopy. Hence, our study enables a robust estimation of plant phenotypes by recognizing small differences of complex cell organelle shapes and their movement, which is beneficial in a comprehensive analysis of the molecular mechanism for cell organelle formation that is independent of technical variations.

Plant phenotyping relevance

顕微鏡画像からERボディの形態・テクスチャ・移動を抽出し、異なる実験条件で頑健性を検証する植物表現型解析手法が中心である。

Code and data availability

The paper's authors publicly release their phenotyping analysis scripts (segmentation, feature extraction, dynamics, clustering) on GitHub and the conda analysis environment on Anaconda Cloud, both with explicit availability statements. Microscope images are said to be in a 'Bioimage database' but no URL is given, so a