Recovering complete plant root system architectures from soil via X-ray μ-Computed Tomography
1 Centre for Plant Integrative Biology, University of Nottingham, Sutton Bonington Campus, Nottingham, LE12 5RD, UK
2 School of Computer Science, University of Nottingham, Jubilee Campus, Nottingham, NG8 1BB, UK
3 School of Biosciences, University of Nottingham, Sutton Bonington Campus, Nottingham, LE12 5RD, UK
Plant Methods 2013, 9:8 doi:10.1186/1746-4811-9-8Published: 20 March 2013
X-ray micro-Computed Tomography (μCT) offers the ability to visualise the three-dimensional structure of plant roots growing in their natural environment – soil. Recovery of root architecture descriptions from X-ray CT data is, however, challenging. The X-ray attenuation values of roots and soil overlap, and the attenuation values of root material vary. Any successful root identification method must both explicitly target root material and be able to adapt to local changes in root properties.
RooTrak meets these requirements by combining the level set method with a visual tracking framework and has been shown to be capable of segmenting a variety of plant roots from soil in X-ray μCT images. The approach provides high quality root descriptions, but tracks root systems top to bottom and so omits upward-growing (plagiotropic) branches.
We present an extension to RooTrak which allows it to extract plagiotropic roots. An additional backward-looking step revisits the previous image, marking possible upward-growing roots. These are then tracked, leading to efficient and more complete recovery of the root system. Results show clear improvement in root extraction, without which key architectural traits would be underestimated.
The visual tracking framework adopted in RooTrak provides the focus and flexibility needed to separate roots from soil in X-ray CT imagery and can be extended to detect plagiotropic roots. The extended software tool produces more complete descriptions of plant root structure and supports more accurate computation of architectural traits.