- Open Access
Rapid phenotyping of crop root systems in undisturbed field soils using X-ray computed tomography
© Pfeifer et al. 2015
- Received: 12 July 2015
- Accepted: 10 August 2015
- Published: 28 August 2015
X-ray computed tomography (CT) has become a powerful tool for root phenotyping. Compared to rather classical, destructive methods, CT encompasses various advantages. In pot experiments the growth and development of the same individual root can be followed over time and in addition the unaltered configuration of the 3D root system architecture (RSA) interacting with a real field soil matrix can be studied. Yet, the throughput, which is essential for a more widespread application of CT for basic research or breeding programs, suffers from the bottleneck of rapid and standardized segmentation methods to extract root structures. Using available methods, root segmentation is done to a large extent manually, as it requires a lot of interactive parameter optimization and interpretation and therefore needs a lot of time.
Based on commercially available software, this paper presents a protocol that is faster, more standardized and more versatile compared to existing segmentation methods, particularly if used to analyse field samples collected in situ. To the knowledge of the authors this is the first study approaching to develop a comprehensive segmentation method suitable for comparatively large columns sampled in situ which contain complex, not necessarily connected root systems from multiple plants grown in undisturbed field soil. Root systems from several crops were sampled in situ and CT-volumes determined with the presented method were compared to root dry matter of washed root samples. A highly significant (P < 0.01) and strong correlation (R2 = 0.84) was found, demonstrating the value of the presented method in the context of field research. Subsequent to segmentation, a method for the measurement of root thickness distribution has been used. Root thickness is a central RSA trait for various physiological research questions such as root growth in compacted soil or under oxygen deficient soil conditions, but hardly assessable in high throughput until today, due to a lack of available protocols.
Application of the presented protocol helps to overcome the segmentation bottleneck and can be considered a step forward to high throughput root phenotyping facilitating appropriate sample sizes desired by science and breeding.
- Non-destructive root phenotyping
- X-ray computed tomography (CT)
- High throughput
- Image analysis
- Field soil
- Root growth dynamics
- Root system architecture (RSA)
- Root thickness
Increasing the throughput for quantitative characterization of plant root system architecture (RSA) is important for plant breeding  and to come to an improved understanding of root–soil interactions [2–4]. In the context of root phenotyping, X-ray computed tomography (CT) has become a powerful tool . Compared to rather classical, destructive methods, in which roots are first washed out of the soil, imaged and then analyzed with commercially available (e.g. WinRHIZO, Regent Instruments Inc., Sainte-Foy, Québec, Canada) or custom-made software (e.g. so-called ‘shovelomics’ approaches as described in [6, 7]), CT encompasses various advantages. In particular, the possibility to follow the same individual root growing over time and to study dynamic root growth and development processes in pot experiments, and, in addition, the opportunity to explore the unaltered configuration of the 3D RSA interacting with a real field soil matrix, makes CT a unique tool for plant research. Moreover, when using destructive methods, in which roots are washed out of the soil, fine roots frequently break off and are washed away, while they can be analyzed in CT scans if the spatial resolution is appropriate .
Yet, using CT, the throughput suffers from the bottleneck of rapid and standardized segmentation methods to extract root structures [4, 9]. Indeed, unaltered RSA can be analyzed by CT, but the segmentation of the root (optical separation of root and soil) is done to a large extent manually and therefore requires a lot of time. Segmentation is usually performed by defining a local threshold for gray values of the CT voxels, classifying them either as root or non-root. These local thresholds often vary throughout a sample due to heterogeneities in the substrate and CT-artefacts. The complete reconstruction of a root system therefore requires a lot of interactive parameter optimization and interpretation, which is supported by software tools that allow identification of connected root systems (e.g. by region growing algorithms as used in [10–12]). Yet, the premise of a completely connected root system can also lead to difficulties. Algorithms that follow root voxels in CT image stacks slice by slice [13, 14] can be confused, if a root seems to be interrupted due to surrounding soil heterogeneities or if roots are only a few voxels in diameter or if multiple root systems contained in soil cores from a field experiment are investigated. These root systems typically consist of cut segments of roots from multiple plants contained in the cylinder of a soil core. Besides this difficulty of segmenting unconnected roots, root segmentation is far more difficult in undisturbed field soils compared to sieved soil filled in pots due to further obstacles. Undisturbed soils frequently contain much higher amounts of organic particles, which are commonly removed in pot experiments by means of sieving the soil. Those organic particles typically have gray values similar to those of roots. Moreover, undisturbed soil samples often show more inhomogeneous moisture distribution compared to sieved and homogenized soils.
Since an ultimate goal of root phenotyping is the characterization of realistic root systems of plant stands in the field , it was the aim of the here presented approach to elaborate a reliable and fast protocol to segment root structures for single or multiple plants, which depends only minimally on the operator and which is applicable for simple root systems, but also for large, arbitrarily complex and unconnected root systems.
Analyzing RSA from field soil can be done with conventional destructive methods too, which are regularly very time-consuming. Indeed, applying so-called ‘shovelomics’ approaches for example, in which RSA traits of maize plants can be visually scored after destructive excavation and washing out from soil, 5–10 min of active working time per sample are needed [6, 7]. Though, shovelomics are not suitable for delicate root systems from field-grown plants (e.g. of cereals), as roots of these plants agglomerate after washing. For delicate root systems of field-grown plants several fundamentally different methods are available, which are all very laborious, as for example WinRHIZO [9, 16] and the profile wall method, originally described by . For application of WinRHIZO, the roots have to be washed even more thoroughly than for shovelomics, imaged and analyzed by the software, which frequently needs, depending on the soil volume to be analyzed, significantly more than 1 h of active working time per sample. Using the profile wall method, several tedious steps have to be performed including the excavation of a walk-in cored hole in the field, preparing the wall in several steps, and counting roots manually using a counting frame. Depending on the number of plants analyzed at the profile wall, several hours or days per profile wall are frequently needed to perform the data acquisition.
The root systems extracted in the framework of the presented segmentation protocol (Fig. 1b, c) showed a high level of complexity and integrity. Consequently, various RSA parameters, such as the number of lateral roots, branching angle or root diameter (Fig. 1c, d) that are interesting for both research and breeding, can be analyzed by the same software used in this work . A highly significant (P < 0.01) and strong correlation (R2 = 0.836) was found for root dry matter of washed root samples and root volumes determined for the same roots with the presented method (Fig. 1e). Given that the CT volume is closely correlated to the fresh weight of the root, the slope of the regression line indicates an average dry weight content of 24 %. This value appears plausible as the samples contained many storage roots, e.g. from Brassica napus and Medicago sativa and storage roots frequently show root dry weight contents between 25 and 35 % . It can be expected that not all roots are identified, both using CT and root washing. However, the correlation shows that the error of missed roots is similar, making the results of both methods comparable.
Previous work has demonstrated the applicability of X-ray computed tomography to investigate effects of various environmental conditions on RSA traits and to study relevant root–soil-interactions, such as effects of soil physical properties on rhizosphere functions [16, 19, 20]. However, in those studies typically disturbed field soil was used, which was sieved to remove organic residues and filled into rather small pots. As plants were sown directly in those pots, the scans usually contained whole root systems. Moreover, the roots were typically resolved in comparatively high resolution due to commonly rather small pot diameters. All these provisions significantly relieve root segmentation. To the knowledge of the authors this is the first study approaching to develop a comprehensive segmentation method suitable for comparatively large columns sampled in situ which contain complex, not necessarily connected root systems from multiple plants grown in undisturbed field soil.
Studies on undisturbed soil samples are needed to increase the general understanding of root growth dynamics, root–soil interactions, root functions and their economic and ecological importance in agro-ecosystems. Available methods for this field of research are still underdeveloped. For instance, virtually all available methods for studying RSA traits of field samples result in a strong underestimation of the proportion of fine roots . Fine roots, defined as roots thinner than 0.2 or 0.5 mm diameter , can make up more than 80 % of the root length of cereals [for a review see 8]. Using the presented protocol, it was observed for the grass–legume mixture samples, scanned at a resolution of 44 µm voxel size, that the major part of the root surface was formed by fine roots thinner than 0.25 mm diameter (Fig. 1c, d; maximum of histogram around 150 µm diameter). Similar observations were made for soil cores taken from wheat plots (data not shown). In future experiments, soil cores with even smaller diameter will be taken at selected positions. In these cores, the smaller voxel dimensions should allow segmentation of even thinner fine roots. Moreover, the use of advanced filtering algorithms, which make advantage of geometrical properties such as width-length or volume-surface ratios to eliminate remaining noise  should be possible in the near future. Those filters may help preserving small and unconnected root segments and could improve the accuracy of the extraction of the root systems, including fine root structures, particularly if other field soils containing higher amounts of organic matter are used. The intended application of refined filtering approaches will then facilitate extraction of further traits of RSA valuable for research and breeding.
Application of the commercially available software VG Studio MAX 2.2 and the two applied add-on modules (‘Coordinate measurement’ and ‘Wall thickness analysis’) in combination with the presented protocol allows for time saving segmentation of arbitrarily complex and unconnected root systems, which can originate from pot experiments or can be collected in situ in the field. For this reason this protocol helps to overcome the segmentation bottleneck and can be considered a step forward to high throughput root phenotyping facilitating appropriate sample sizes desired by science and breeding. Moreover, a fast and simple way for a quantitative determination of root thickness distribution, which is an important but normally only very tediously determinable phenotypic trait, has been applied thereby. Further RSA parameters interesting for both research and breeding can be analyzed by the same software used here. Therefore, the application of the specified software in combination with the described protocol will result in a significant progress for a large spectrum of future studies performed in the field of crop phenotyping.
Plant material and collection of undisturbed soil samples
Soil and sample properties and scanning parameters for X-ray computed tomography
Pot experiment (controlled conditions)
Samples collected in situ in the field
Samples collected in situ in the field
Samples collected in situ in the field
Klein-Altendorf: Haplic Luvisol, sieved
Eschikon: Eutric Cambisol; Reckenholz: Pseudogleyed Cambisol
Eschikon: Eutric Cambisol
Reckenholz: Pseudogleyed Cambisol
Plant age (months)
Cylinder internal diameter (cm)
3.4 (small samples) and 10 (large samples)
Number of plants per cylinder
Height of scanned part of root system (cm)
15 (small samples) and 10 (large samples)
Height of analyzed part of root system (cm)
15 (small samples) and 10 (large samples)
Voxel size (mm)
0.120 (large samples) and 0.044 (small samples)
1 × 1
2 × 2
2 × 2
2 × 2
450 (large samples) and 350 (small samples)
Number of images per subscan
0.1 mm copper
0.1 mm copper
0.4 mm copper
0.4 (large samples) and 0.1 mm (small samples) copper
Observation ROI option
Exposure time per image (ms)
131 (small samples) and 1000 (large samples)
Scan duration (min)
Multiscan and number of subscans
Small samples: yes (3), large samples: no
Downscaling to unsigned 16 bit
Auto scan optimizer
Beam hardening correction
Assuming different materials, value 4
Assuming different materials, value 3.6
Assuming different materials, value 3.6
Assuming different materials, value 3.6
X-ray computed tomography scans were performed at the Swiss Federal Institute of Technology Zurich (ETH Zürich, Switzerland) using a phoenix v|tome|x s 240 X-ray scanner (GE Sensing & Inspection Technologies GmbH, Wunstorf, Germany). Two different configurations of acquisition parameters for tomography were chosen for the two different sample sizes (Table 1). Volumes were reconstructed using the software datos|x (GE Sensing & Inspection Technologies GmbH, Wunstorf, Germany). For reconstruction (in 32-bit float) an auto-scanoptimization and a beam hardening correction were performed. In case of multiscans, the single subscans were combined while the data were reconstructed. For this reason, data from multiscans could be analyzed together. Very slight gray value differences could be observed in the transitions from one subscan to the other. However, for the subsequent analysis it was not necessary to normalize for those differences in order to achieve seamless transitions.
Frequency distributions of root diameters were determined using the add-on module ‘Wall thickness analysis’ of the VG Studio MAX 2.2 software. Similar to the advanced surface determination algorithm the diameter of the root was calculated along the root surface (Fig. 1c, d). The surface determined by the ‘Advanced surface determination’-tool served as the starting contour. Several parameters and tolerance values can be manually adjusted using the ‘Wall thickness analysis’-tool, such as minimum and maximum thickness of the target structure and operating search angles.
In order to compare the presented protocol with another common method to extract root systems from CT volumes [10–12] the region growing tool from VG Studio MAX 2.2 was chosen (Fig. 1a). The tool is based on a region growing algorithm starting at manually selected seed points (here root material). As the local thresholds for segmenting roots commonly vary throughout the sample (due to heterogeneities in the substrate and CT-artefacts), it was, even though the adaptive mode was used, not possible to apply the algorithm to the entire CT volume. Instead, the search area of the algorithm needed to be restricted in 3D, which made it necessary to apply the algorithm in a plenty of single steps (by determining new seed points). In each step the newly segmented root structure was added to the already segmented part of the root system (saved as a ROI), and in each step the tolerance value needed to be adjusted manually to the local threshold.
JP, NK, TC and AW designed the experiments. JP and TC collected the samples and carried out the CT measurements. JP, TC and NK analyzed the data. AW and NK supervised the study. JP and AW drafted the manuscript. All authors read, ameliorated and approved the final manuscript.
Compliance with ethical guidelines
Competing interests The authors declare that they have no competing interests.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Walter A, Liebisch F, Hund A (2015) Plant phenotyping: from bean weighing to image analysis. Plant Methods 11:14PubMed CentralView ArticlePubMedGoogle Scholar
- Lynch JP, Brown KM (2012) New roots for agriculture: exploiting the root phenome. Philos Trans R Soc B 367:1598–1604View ArticleGoogle Scholar
- Gregory PJ, Atkinson CJ, Bengough AG, Else MA, Fernández-Fernández F, Harrison RJ, Schmidt S (2013) Contributions of roots and rootstocks to sustainable, intensified crop production. J Exp Bot 64:1209–1222View ArticlePubMedGoogle Scholar
- Haling RE, Tighe MK, Flavel RJ, Young IM (2013) Application of X-ray computed tomography to quantify fresh root decomposition in situ. Plant Soil 372:619–627View ArticleGoogle Scholar
- Mooney SJ, Pridmore TP, Helliwell J, Bennett MJ (2012) Developing X-ray computed tomography to non-invasively image 3-D root systems architecture in soil. Plant Soil 352:1–22View ArticleGoogle Scholar
- Trachsel S, Kaeppler SM, Brown KM, Lynch JP (2011) Shovelomics: high throughput phenotyping of maize (Zea mays L.) root architecture in the field. Plant Soil 341:75–87View ArticleGoogle Scholar
- Colombi T, Kirchgessner N, Le Marié CA, York L, Lynch J, Hund A (2015) Next generation shovelomics: set up a tent and REST. Plant Soil 388:1–20View ArticleGoogle Scholar
- Pierret A, Moran CJ, Doussan C (2005) Conventional detection methodology is limiting our ability to understand the roles and functions of fine roots. New Phytol 166:967–980View ArticlePubMedGoogle Scholar
- Metzner R, Eggert A, van Dusschoten D, Pflugfelder D, Gerth S, Schurr U, Uhlmann N, Jahnke S (2015) Direct comparison of MRI and X-ray CT technologies for 3D imaging of root systems in soil: potential and challenges for root trait quantification. Plant Methods 11:17PubMed CentralView ArticlePubMedGoogle Scholar
- Flavel RJ, Guppy CN, Tighe M, Watt M, McNeill A, Young IM (2012) Non-destructive quantification of cereal roots in soil using high-resolution X-ray tomography. J Exp Bot 63:2503–2511View ArticlePubMedGoogle Scholar
- Pfeifer J, Kirchgessner N, Walter A (2014) Artificial pores attract barley roots and can reduce artifacts of pot experiments. J Plant Nutr Soil Sci 177:903–913View ArticleGoogle Scholar
- Koebernick N, Weller U, Huber K, Schlüter S, Vogel HJ, Jahn R, Vereecken H, Vetterlein D (2014) In situ visualization and quantification of three dimensional root system architecture and growth using X-ray computed tomography. Vadose Zone J 13:8View ArticleGoogle Scholar
- Mairhofer S, Zappala S, Tracy S, Sturrock C, Bennett MJ, Mooney SJ, Pridmore TP (2012) RooTrak: automated recovery of three-dimensional plant root architecture in soil from X-ray microcomputed tomography images using visual tracking. Plant Physiol 158:561–569PubMed CentralView ArticlePubMedGoogle Scholar
- Mairhofer S, Zappala S, Tracy S, Sturrock C, Bennett MJ, Mooney SJ, Pridmore TP (2013) Recovering complete plant root system architectures from soil via X-ray μ-computed tomography. Plant Methods 9:8PubMed CentralView ArticlePubMedGoogle Scholar
- Araus JL, Cairns JE (2013) Field high-throughput phenotyping: the new crop breeding frontier. Trends Plant Sci 19:52–61View ArticlePubMedGoogle Scholar
- Tracy SR, Black CR, Roberts JA, Dodd IC, Mooney SJ (2015) Using X-ray computed tomography to explore the role of abscisic acid in moderating the impact of soil compaction on root system architecture. Environ Exp Bot 110:11–18View ArticleGoogle Scholar
- Böhm W (1976) In situ estimation of root length at natural soil profiles. J Agric Sci 87:365–368View ArticleGoogle Scholar
- Kawano K, Fukuda WMG, Cenpukdee U (1987) Genetic and environmental effects on dry matter content of cassava root. Crop Sci 27:69–74View ArticleGoogle Scholar
- Tracy SR, Black CR, Roberts JA, Sturrock C, Mairhofer S, Craigon J, Mooney SJ (2012) Quantifying the impact of soil compaction on root system architecture in tomato (Solanum lycopersicum L.) by X-ray micro-computed tomography (CT). Ann Bot 110:511–519PubMed CentralView ArticlePubMedGoogle Scholar
- Daly KR, Mooney SJ, Bennett MJ, Crout NMJ, Roose T, Tracy SR (2014) Assessing the influence of the rhizosphere on soil hydraulic properties using X-ray computed tomography and numerical modelling. J Exp Bot 66:2305–2314View ArticleGoogle Scholar