Automated image analyses for separating plant roots from soil debris elutrated from soil cores

Historically, destructive root sampling has been labor intensive and requires manual separation of extraneous organic debris recovered along with the hydropneumatic elutriation method of separating plant roots from soils. Quantification of root system demographics by public domain National Institute of Health (NIH-Image) and Root Image Processing Laboratory (RIPL) image processing algorithms has eliminated much of the labor-intensive manual separation. This was accomplished by determining the best length to diameter ratio for each object during image analyses. Objects with a length to diameter ratio less than a given threshold are considered non-root materials and are rejected automatically by computer algorithms. Iterative analyses of length to diameter ratios showed that a 15:1 ratio was best for separating images of maize (Zea mays L.) roots from associated organic debris. Using this threshold ratio for a set of 24 soil cores, a highly significant correlation (r² = 0.89) was obtained between computer image processed total root length per core and actual root length. A linear relationship (r² = 0.80) was observed between root lengths determined by NIH-Image analyses and lengths determined independently by the RIPL imaging system, using the same maize root + debris samples. This correlation demonstrates that computer image processing provides opportunities for comparing root length parameters between different laboratories for samples containing debris.     

A method to separate plant roots from soil and analyze root surface area

Analysis of the effects of soil management practices on crop production requires knowledge of these effects on plant roots. Much time is required to wash plant roots from soil and separate the living plant roots from organic debris and previous years’ roots. We developed a root washer that can accommodate relatively large soil samples for washing. The root washer has a rotary design and will accommodate up to 24 samples (100 mm diam. by 240 mm long) at one time. We used a flat-bed scanner to digitize an image of the roots from each sample and used a grid system with commercially-available image analysis software to analyze each sample for root surface area. Sensitivity analysis and subsequent comparisons of ‘dirty’ samples containing the roots and all the organic debris contained in the sample and ‘clean’ samples where the organic debris was manually removed from each sample showed that up to 15% of the projected image could be coveredwith debris without affecting accuracy and precision of root surface area measurements. Samples containing a large amount of debris may need to be partitioned into more than one scanning tray to allow accurate measurements of the root surface area. Sample processing time was reduced from 20 h, when hand separation of roots from debris was used, to about 0.5 h, when analyzing the image from an uncleaned sample. The method minimizes the need for preprocessing steps such as dying the roots to get better image contrast for image analysis. Some information, such as root length, root diameter classes and root weights, is not obtained when using this technique. Root length measurements, if needed, could be made by hand on the digital images. Root weight measurement would require sample cleaning and the advantage of less processing time per sample with this method would be lost. The significance of the tradeoff between information not obtained using this technique and the ability to process a greater number of samples with the time and personnel resources available must be determined by the individual researcher and research objectives.     

Root length and diameter measurement using NIH Image: application of the line-intercept principle for diameter estimation

The objective of this study was to develop an image analysis algorithm for estimating the length versus diameter distribution of washed root samples. Image analysis was performed using a Macintosh computer and the public domain NIH Image program. After an appropriate binary image of roots was obtained, the image was processed to get the thinned image to calculate the length of the roots. The edge pixel of the binary image was then deleted and root length was calculated again. This `edge deletion–length calculation' cycle was repeated until no root pixel was left in the image. Repeated edge deletion removed one pixel layer from around the periphery of root objects in each iteration. The number of edge deletions, which is equivalent to the intercept length, can be used to estimate the root diameter. We used the vertical or horizontal intercept length, whichever was shorter. The accuracy of diameter estimation due to orientation of objects varied from 89.1 to 126.0%. Branching root systems consist of several orders of laterals, and as the root branches to a higher order, the diameter of the roots becomes smaller. Therefore, edge deletions eliminate sequentially from the highest order roots, which have the smallest diameter, to the lowest order roots, which have the widest diameter. Thus, the length and diameter of each root order can be calculated by the proposed method. For verification, images of copper wire of 0.23, 0.50, and 1.0 mm diameter were analyzed. The results showed reasonable agreement with the expected distribution of length versus diameter for randomly oriented objects, and consequently the wire length of each diameter could be estimated. The proposed method was tested for primary and secondary roots of water-cultured rice (Oryza sativa L.), and it was proven that the method can provide accurate length and diameter measurements for each root order.     

A color composite technique for detecting root dynamics of barley (Hordeum vulgare L.) from minirhizotron images

Quantification of root dynamics by destructive methods is confounded by high coefficients of variation and loss of fine roots. The minirhizotron technique is non-destructive and allows for sequential root observations to be made at the same depth in situ. Observations can be stored on video tape which facilitates data handling and computer-aided image processing. A color composite technique using digital image analyses was adapted in this study to detect barley root dynamics from sequential minirhizotron images. Plants were grown in the greenhouse in boxes (80 × 80 × 75 cm) containing soil from a surface horizon of a Typic Cryoboroll. A minirhizotron was installed at a 45°C angle in each box. Roots intersecting the minirhizotron were observed and video-recorded at tillering, stem extension, heading, dough and ripening growth stages. The images from a particular depth were digitized from the analog video then registered to each other. Discrimination of roots from the soil matrix gave quantitative estimates of root appearance and disappearance. Changes in root appearance and disappearance were detected by assigning a separate primary color (red, green, blue) to selected growth stages, then overlaying the images to create red-green and red-green-blue color composites. The resulting composites allowed for a visual interpretation and quantification of barley root dynamics in situ.     

An improved method for clearing and staining free-hand sections and whole-mount samples

BACKGROUND AND AIMS: Free-hand sectioning of living plant tissues allows fast microscopic observation of internal structures. The aim of this study was to improve the quality of preparations from roots with suberized cell walls. A whole-mount procedure that enables visualization of exo- and endodermal cells along the root axis was also established. METHODS: Free-hand sections were cleared with lactic acid saturated with chloral hydrate, and observed with or without post-staining in toluidine blue O or aniline blue. Both white light and UV light were used for observation. Lactic acid was also used as a solvent for berberine, and fluorol yellow for clearing and staining the samples used for suberin observation. This procedure was also applied to whole-mount roots with suberized celllayers. KEY RESULTS: Clearing of sections results in good image quality to observe the tissue structure and cell walls compared with non-cleared sections. The use of lactic acid as a solvent for fluorol yellow proved superior to previously used solvents such as polyethylene glycol-glycerol. Clearing and fluorescence staining of thin roots such as those of Arabidopsis thaliana were successful for suberin visualization in endodermal cells within whole-mount roots. For thicker roots, such as those of maize, sorghum or tea, this procedure could be used for visualizing the exodermis in a longitudinal view. Clearing and staining of peeled maize root segments enabled observation of endodermal cell walls. CONCLUSIONS: The clearing procedure using lactic acid improves the quality of images from free-hand sections and clearings. This method enhances the study of plant root anatomy, in particular the histological development and changes of cell walls, when used in combination with fluorescence microscopy.     

Analysis of root images from auger sampling with a fast procedure: a case of application to sugar beet

Manual line-intersect methods for estimating root length are being progressively replaced by faster and more accurate image analysis procedures. These methods even allow the estimation of some more root parameters (e.g., diameter), but still require preliminary labour-intensive operations. Through a task-specific macro function written in a general-purpose image analysis programme (KS 300 – Zeiss), the processing time of root images was greatly reduced with respect to skeletonisation methods by using a high-precision algorithm (Fibrelength). This has been previously proposed by other authors, and estimates length as a function of perimeter and area of the digital image of roots. One-bit binary images were acquired, aiming at large savings in computer memory, and automatic discrimination of roots against extraneous objects based on their elongation index (perimeter²/area), was performed successfully. Of four tested spatial resolutions (2.9, 5.9, 8.8, 11.8 pixel mm^–1), in clean samples good accuracy in root length estimation was achieved at 11.8 pixel mm^–1, up to a root density of 5 cm cm^–2 on the scanner bed. This resolution is theoretically suitable for representing roots at least 85 m wide. When dealing with uncleaned samples, a thick layer of water was useful in speeding up spreading of roots on the scanner bed and avoiding underestimation of their length due to overlaps with organic debris. A set of fibrous root samples of sugar beet (Beta vulgaris var. saccharifera L.) collected at harvest over two years at Legnaro (NE Italy) was analysed by applying the above procedure. Fertilisation with 100 kg ha^–1 of nitrogen led to higher RLD (root length density in soil) in shallow layers with respect to unfertilised controls, whereas thicker roots were found deeper than 80 cm of soil without nitrogen.     

Total root length as estimated from small sub-samples

A method is proposed for estimating the total length of a root system from sub-samples. This method is based on the measurement of the length and diameter of small pieces of roots, and on the measurements of the bulk density of root sub-samples. It is assumed that roots are cylinders with a given bulk density. The length and diameter of small root pieces are measured by image analysis. A weighted quadratic mean (W.Q.M.) root diameter is then calculated and used in estimating the root length. This W.Q.M. diameter is defined as the real mean diameter of an equivalent single root with the same length and volume as the tested root system. The accuracy of prediction is demonstrated for one theoretical root system. The standard deviation of estimation can be calculated using sampling simulations.     

Automatic discrimination of fine roots in minirhizotron images

Minirhizotrons provide detailed information on the production, life history and mortality of fine roots. However, manual processing of minirhizotron images is time-consuming, limiting the number and size of experiments that can reasonably be analysed. Previously, an algorithm was developed to automatically detect and measure individual roots in minirhizotron images. Here, species-specific root classifiers were developed to discriminate detected roots from bright background artifacts. Classifiers were developed from training images of peach (Prunus persica), freeman maple (Acer x freemanii) and sweetbay magnolia (Magnolia virginiana) using the Adaboost algorithm. True- and false-positive rates for classifiers were estimated using receiver operating characteristic curves. Classifiers gave true positive rates of 89-94% and false positive rates of 3-7% when applied to nontraining images of the species for which they were developed. The application of a classifier trained on one species to images from another species resulted in little or no reduction in accuracy. These results suggest that a single root classifier can be used to distinguish roots from background objects across multiple minirhizotron experiments. By incorporating root detection and discrimination algorithms into an open-source minirhizotron image analysis application, many analysis tasks that are currently performed by hand can be automated.     

A comparison between minirhizotron and monolith sampling methods for measuring root growth of maize (Zea mays L.)

Transparent plastic minirhizotron tubes have been used to evaluate spatial and temporal growth activities of plant root systems. Root number was estimated from video recordings of roots intersecting minirhizotron tubes and of washed roots extracted from monoliths of the same soil profiles at the physiological maturity stage of a maize (Zea mays L.) crop. Root length was measured by the line intercept (LI) and computer image processing (CIP) methods from the monolith samples.There was a slight significant correlation (r=0.28, p<0.005) between the number of roots measured by minirhizotron and root lengths measured by the LI method, however, no correlation was found with the CIP method. Using a single regression line, root number was underestimated by the minirhizotron method at depths between 0–7.6 cm. A correlation was found between root length estimated by LI and CIP. The slope of estimated RLD was significant with depth for these two methods. Root length density (RLD) measured by CIP showed a more erratic decline with distance from the plant row and soil surface than the LI method.     

X-ray absorption and phase contrast imaging to study the interplay between plant roots and soil structure

Plant performance is, at least partly, linked to the location of roots with respect to soil structure features and the micro-environment surrounding roots. Measurements of root distributions from intact samples, using optical microscopy and field tracings have been partially successful but are imprecise and labour-intensive. Theoretically, X-ray computed micro-tomography represents an ideal solution for non-invasive imaging of plant roots and soil structure. However, before it becomes fast enough and affordable or easily accessible, there is still a need for a diagnostic tool to investigate root/soil interplay. Here, a method for detection of undisturbed plant roots and their immediate physical environment is presented. X-ray absorption and phase contrast imaging are combined to produce projection images of soil sections from which root distributions and soil structure can be analyzed. The clarity of roots on the X-ray film is sufficient to allow manual tracing on an acetate sheet fixed over the film. In its current version, the method suffers limitations mainly related to (i) the degree of subjectivity associated with manual tracing and (ii) the difficulty of separating live and dead roots. The method represents a simple and relatively inexpensive way to detect and quantify roots from intact samples and has scope for further improvements. In this paper, the main steps of the method, sampling, image acquisition and image processing are documented. The potential use of the method in an agronomic perspective is illustrated using surface and sub-surface soil samples from a controlled wheat trial. Quantitative characterization of root attributes, e.g. radius, length density, branching intensity and the complex interplay between roots and soil structure, is presented and discussed. This revised version was published online in June 2006 with corrections to the Cover Date.     

Accurate root length and diameter measurement using NIH Image: use of Pythagorean distance for diameter estimation

This report describes an image analysis algorithm to estimate the length versus diameter of washed root samples accurately. Image analysis was performed using a Macintosh computer and the public domain NIH Image program. The binary image of the roots was processed to get the thinned image to calculate the length of the roots. The pixels of the root in a binary image were then stripped off from around the periphery based on the pixel's Pythagorean distance from the nearest background pixel. The length of the remaining root in each stripping off process was calculated after the image was thinned. Images (300 dpi) of copper wire of 0.23, 0.5, 1.0 mm diameter were analyzed for verification of the usefulness of the procedure. The results showed that more than 93% of the wires in each diameter wire were calculated to be in diameter classes including the true diameter and its adjoining classes: 93.6% of the wires of 0.23 mm diameter appeared in the 0.098–0.38 mm diameter classes, 96.19% of the wires of 0.5 mm diameter appeared in the 0.38–0.61 mm diameter classes, and 96.17% of the wires of 1 mm diameter appeared in the 0.85–1.08 mm diameter classes. The proposed method was tested for primary and secondary roots of water-cultured rice (Oryza sativa L.) and it was proven that the method could provide accurate length and diameter measurements for each root order. In addition, it was found that the method could provide the lengths of the thick primary, thin primary, and secondary roots. The effectiveness of applying sharpening for the grayscale image before making the binary image is also discussed.     

Using digital image analysis to quantify the architectural parameters of roots grown in thin rhizotrons

Abstract

Roots grown in small chambers or rhizotrons combined with image analysis could be effective for screening traits of root morphology and architecture prior to field experimentation. The objective of this study was to evaluate and compare digital images of root systems with conventional root scans (RSCN) for several sorghum varieties grown in rhizotrons. Plants were grown in duplicate in slim (1.5‐cm thick) rhizotrons angled 15° to the vertical, filled with sandy topsoil and subsoil at field capacity (FC) or 50% available water content (AWC) in a split‐plot design in a glasshouse (16/24°C). Root growth visible on the glass surface was recorded weekly, and roots were washed out 42 and 90 days after sowing for digital imaging of the root systems on pin boards (PBI). The roots were then sub‐sampled to quantify the baseline root parameters from digital RSCN using WinRhizo® software with a flat‐bed scanner. Rhizotron images revealed little of the root system (1–2%) but PBI “recovered” up to 70% of the total root length (TRL) measured by RSCN. A high a priori image contrast, increased image resolution (9.1–12 megapixels) and optimized contrast threshold between roots and background were the key parameters in quantifying the roots from digital images. PBI revealed significant differences in rooting patterns, especially distribution of root length density (RLD) in the profile, which is useful for selecting varieties with improved resource capture. However, the root diameters estimated in the PBI were significantly larger than those measured by RSCN due to lack of contrast, root clustering and overlay.     

荧光粉激发型LED光对拟南芥生长发育的影响

在室内人工光源照射条件下, 探究150 μmol∙m^-2·s^-1光强下不同比例的红蓝光对拟南芥(Arabidopsis thaliana)生长发育的影响。以哥伦比亚(Columbia-0)野生型拟南芥为研究对象, 采用荧光粉激发型LED作为植物生长光源, 以SrSiAlN₃为红色基底, 调节荧光粉添加量获得不同红蓝光谱, 考察不同光照条件下拟南芥萌发率、根长、株高、叶绿素含量和相对电导率等参数的变化规律。结果表明, 在荧光粉激发型LED光照调节下的拟南芥具有更高的萌发率、根长、株高、叶绿素含量和相对电导率, 且在红蓝光质比为2:1时萌发率(95.63%)和叶绿素含量(26.7)最高; 在红蓝光质比为4:5时根长(4.19 cm)较长; 在红蓝光质比为4:1时株高(15.5 cm)较高; 在红蓝光质比为4:5时相对电导率(40.5 S·m^-1)较大。研究结果表明相对平衡的光质(红蓝光质比为4:1)有利于拟南芥生长发育, 且减少蓝光比例对根系生长及叶绿素积累有一定的促进作用。研究表明不同光谱对模式植物拟南芥的生长发育有较明显的影响, 改变光谱组成可以对植物的生长发育起不同程度的调控作用。     

Recording,processing and analysis of grass root images from a rhizotron

To develop and test a system for computer-assisted image analysis, repeated video recordings of reed canary-grass roots (Phalaris arundinacea L.) were made in an 18-window rhizotron. The images were digitized and processed using a Unix computer and the Khoros software development environment.Two image sizes, 126×95 mm and 61×46 mm, both comprising 650 × 490 pixels, were compared. Among image processing techniques used were median filtering, segmentation and skeletonization. Root area and length in both the topsoil and subsoil were estimated using the two image sizes. The resolution (image size) strongly affected the calculated root lengths. The results were compared with root length measurements obtained manually.Statistically significant differences in root length and area in the topsoil were detected between the sampling dates using the computer-assisted methods. Possible sources of error and methods for reducing them are discussed.     

Accurate root length measurement by image analysis

Algorithms for estimating root length by image analysis should lead to results that have no systematic error (bias), be insensitive to preferential root orientation, valid across a wide range of sample sizes and adjust for overlap between roots in samples, to reduce the effort needed in spreading out root systems. We propose a new algorithm that forms a compromise between small bias and robustness (insensitivity to variation in sample size and preferential root orientation), and provide a simple way of dealing with root overlap. Image analysis was performed on a Macintosh computer using the public domain NIH Image program. The digital image of the root was processed to get the thinned image (skeleton). The numbers of orthogonally and diagonally connected pairs of pixels (N _o and N _d, respectively) in the skeleton were counted separately and used for length (L) calculation. A new length calculation equation was introduced so that the effect of orientation on length calculation was minimized; L=[N d ²+(N _d+N _o/2)²]^1/2+N _o/2. The maximum error due to orientation of a single line was evaluated for an ideal line, and the analysis revealed that the new equation was less affected by orientation than previous equations. Copper wire and rice (Oryza sativa L.) roots containing both primary and fine secondary root were measured manually and with image analysis. The two methods showed good agreement within 1.5%. The proposed image analysis method yielded length estimates with CV from 0.23 to 0.88%, which was lower than the CVs of the line-intersect method. Moreover, the lengths of overlapping samples were calculated correctly because the image analysis method distinguished an overlapping pixel from a thinned image, while the calculation with the line-intersect method showed underestimation because overlaps were not considered in that method. This revised version was published online in June 2006 with corrections to the Cover Date.     

Root sampling methods - applications and limitations of the minirhizotron technique

Applications and limitations of the minirhizotron technique (non-destructive) in relation to two frequently used destructive methods (soil coreing and ingrowth cores) is discussed. Sequential coreing provides data on standing crop but it is difficult to obtain data on root biomass production. Ingrowth cores can provide a quick estimate of relative fine-root growth when root growth is rapid. One limitation of the ingrowth core is that no information on the time of ingrowth and mortality is obtained.The minirhizotron method, in contrast to the destructive methods permits simultaneous calculation of fine-root length production and mortality and turnover. The same fine-root segment in the same soil space can be monitored for its life time, and stored in a database for processing. The methodological difficulties of separating excavated fine roots into living and dead vitality classes are avoided, since it is possible to judge directly the successive ageing of individual roots from the images. It is concluded that the minirhizotron technique is capable of quantifying root dynamics (root-length production, mortality and longevity) and fine-root decomposition. Additionally, by combining soil core data (biomass, root length and nutrient content) and minirhizotron data (length production and mortality), biomass production and nutrient input into the soil via root mortality and decomposition can be estimated.     

Conventional detection methodology is limiting our ability to understand the roles and functions of fine roots

We lack a thorough conceptual and functional understanding of fine roots. Studies that have focused on estimating the quantity of fine roots provide evidence that they dominate overall plant root length. We need a standard procedure to quantify root length/biomass that takes proper account of fine roots. Here we investigated the extent to which root length/biomass may be underestimated using conventional methodology, and examined the technical reasons that could explain such underestimation. Our discussion is based on original X-ray-based measurements and on a literature review spanning more than six decades. We present evidence that root-length recovery depends strongly on the observation scale/spatial resolution at which measurements are carried out; and that observation scales/resolutions adequate for fine root detection have an adverse impact on the processing times required to obtain precise estimates. We conclude that fine roots are the major component of root systems of most (if not all) annual and perennial plants. Hence plant root systems could be much longer, and probably include more biomass, than is widely accepted.     

An image analysis method for determination of spatial colonization patterns of bacteria in plant rhizosphere

A method that allows the rapid visualization of bacterial spatial colonization patterns on roots for the determination of general colonization trends was developed. This method, which analyzes images of roots, and bioluminescence-enhanced images of bacterial colonization patterns on these roots, was used to study the colonization patterns of seed-applied Enterobacter cloacae strain E6 on 3-day-old cucumber plants. Conventional dilution-plating methods indicated that E6 colonized cucumber tap roots in high populations and that these populations significantly decreased as the distance from the seed increased. In addition to confirming these observations, image analysis indicated that colonization by E6 significantly decreased on lateral roots as the distance increased horizontally away from the tap root, and that this bacterium did not evenly cover the most densely colonized regions of the cucumber root system. Results from these experiments indicate that the majority of E6 populations on cucumber roots after seed application are limited to the upper regions of the tap root and that E6 does not effectively colonize other regions of the root system. Received: 15 June 1988 / Received revision: 19 November 1998 / Accepted: 29 November 1998     

Automatic image recognition to determine morphological development and secondary metabolite accumulation in hairy root networks

This study focuses on the morphological development and secondary metabolite production of the red pigments from the group of betacyanins in hairy roots of Beta vulgaris. We demonstrate a working, medium throughput, customized, automatic image recognition solution for hairy roots on agar plates including the evaluation of 12 experimental samples. Image acquisition is conducted under comparable para‐meters using a tripod with light emitting diode background lighting and a digital single lens reflex camera. The server‐based image recognition system developed together with Wimasis GmbH, Munich, Germany helps to obtain not only quantitative values for morphological parameters, such as segment lengths and widths or metabolite concentrations, but also global parameters of root growth, such as total root length or the number of branching points. Using timed diagrams the development of the total root length, the total number of branching points, and the mean pigment concentration during the cultivation period were determined. The generated data present the basis for detailed mathematical modeling in order to achieve a structured growth model for hairy roots. A mathematical model for growth of hairy roots can be used to decrease experimental efforts as well as to optimize cultivation conditions and the bioreactor design.     

Distinguishing between metabolically active and inactive roots by combined staining with 2,3,5-triphenyltetrazolium chloride and image colour analysis

The objective of the present work was to develop a method to distinguish between metabolically inactive and active parts of plant roots. White clover (Trifolium repens L.) roots were stained with 2,3,5-triphenyltetrazolium chloride (TTC) followed by root colour classification with an interactive scanner-based image analysis programme (WinRHIZO). Roots inactivated by boiling were unstained and pale brown, whereas fresh samples with predominantly metabolically active roots turned dark red, red or pale red after staining. A small amount of very young, presumable active roots (0.8% of total active root length) failed to stain red with TTC. The colour analysis of inactive and active roots was based on four colour classes for boiled roots and seven classes for fresh roots, respectively, as defined upon visual examination of images. Pixel colours falling outside the defined classes were allocated to the nearest defined class – an option that increased objectivity and stability and reduced the required number of colour classes. For the fresh white clover roots, 75–86% of the total root length was determined as active, while 3–7% of the boiled roots fell into the same category. The percentage of total root length measured by WinRHIZO that was identified as metabolically active was linearly correlated with the percentage of fresh roots in mixtures of fresh and boiled roots (R²=0.99). Colour classes chosen à priori from one experiment could be used to distinguish fairly satisfactorily between active and inactive roots of another white clover cultivar grown under other conditions, but failed to classify activity in ryegrass (Lolium multiflorum Lam.) root samples. In the latter case, colour classes needed to be re-defined in order to produce reliable data. Our work shows that WinRHIZOs colour identification sub-module provides a new promising tool to classify root activity as identified after staining with TTC, but colour classes must be carefully evaluated on every new occasion.