An Improved Method for Automatic Segmentation of Fine Roots in Soil from X-ray Computed Tomography Images

Roots are the hidden half of plants as they grow in soil, a non-transparent medium. X-ray computed tomography (CT) makes it possible to visualize and quantify root growth in situ. Automatic segmentation algorithms are challenged by (1) the overlap of grey value histograms for roots and water filled pores (Mooney et al. 2012) and (2) by the fact that individual root segments are only a few voxels wide in CT scans of typical pot sizes required for unconstrained root growth in early stages of plant development. In this study, we developed a macro for automatic segmentation of fine roots in soil from X-ray CT images using the open source software ImageJ (Fig. 1). Linear correlation between root length determined from destructive analysis and X-ray were highly correlated with a coefficient of determination (R²) of 0.92. Compared with the previous methods of region growing in VG Studio (Flavel et al., 2012) and Root1 (Flavel et al., 2017), the protocol in this study could extract roots more automatically and precisely. This precise detection of roots as thin as 90 µm in its original 3D spatial context allows for an unprecedented analysis of undisturbed root system architecture down to the fine root level.

Fig.1 Steps of roots segmentation. (A) Original image. (B) Denoising and inverting. (C-F) Detection of tubular roots after different Gaussian filters (G) Binarized image after hysteresis thresholding of image C-F. (H) Connected root cluster. (I) 3D root network.

References:

Mooney et al. (2012) Developing X-ray computed tomography to non-invasively image 3-D root systems architecture in soil. Plant Soil, 352: 1-22.

Flavel et al. (2012) Non-destructive quantification of cereal roots in soil using high-resolution X-ray tomography. J Exp Bot 63: 2503-2511.

Flavel et al. (2017) An image processing and analysis tool for identifying and analysing complex plant root systems in 3D soil using non-destructive analysis: Root1. PloS one 12(5): e0176433.