Semi-automated Root Image Analysis (saRIA)

Cyber-agricultural systems for crop breeding and sustainable production

The cyber-agricultural system (CAS) represents an overarching framework of agriculture that leverages recent advances in ubiquitous sensing, artificial intelligence, smart actuators, and scalable cyberinfrastructure (CI) in both breeding and production agriculture. We discuss the recent progress and perspective of the three fundamental components of CAS - sensing, modeling, and actuation - and the emerging concept of agricultural digital twins (DTs). We also discuss how scalable CI is becoming a key enabler of smart agriculture. In this review we shed light on the significance of CAS in revolutionizing crop breeding and production by enhancing efficiency, productivity, sustainability, and resilience to changing climate. Finally, we identify underexplored and promising future directions for CAS research and development.

Semantic segmentation of plant roots from RGB (mini-) rhizotron images-generalisation potential and false positives of established methods and advanced deep-learning models

BACKGROUND

Manual analysis of (mini-)rhizotron (MR) images is tedious. Several methods have been proposed for semantic root segmentation based on homogeneous, single-source MR datasets. Recent advances in deep learning (DL) have enabled automated feature extraction, but comparisons of segmentation accuracy, false positives and transferability are virtually lacking. Here we compare six state-of-the-art methods and propose two improved DL models for semantic root segmentation using a large MR dataset with and without augmented data. We determine the performance of the methods on a homogeneous maize dataset, and a mixed dataset of > 8 species (mixtures), 6 soil types and 4 imaging systems. The generalisation potential of the derived DL models is determined on a distinct, unseen dataset.

RESULTS

The best performance was achieved by the U-Net models; the more complex the encoder the better the accuracy and generalisation of the model. The heterogeneous mixed MR dataset was a particularly challenging for the non-U-Net techniques. Data augmentation enhanced model performance. We demonstrated the improved performance of deep meta-architectures and feature extractors, and a reduction in the number of false positives.

CONCLUSIONS

Although correction factors are still required to match human labelled root lengths, neural network architectures greatly reduce the time required to compute the root length. The more complex architectures illustrate how future improvements in root segmentation within MR images can be achieved, particularly reaching higher segmentation accuracies and model generalisation when analysing real-world datasets with artefacts-limiting the need for model retraining.

Integrated phenotyping of root and shoot growth dynamics in maize reveals specific interaction patterns in inbreds and hybrids and in response to drought

In recent years, various automated methods for plant phenotyping addressing roots or shoots have been developed and corresponding platforms have been established to meet the diverse requirements of plant research and breeding. However, most platforms are only either able to phenotype shoots or roots of plants but not both simultaneously. This substantially limits the opportunities offered by a joint assessment of the growth and development dynamics of both organ systems, which are highly interdependent. In order to overcome these limitations, a root phenotyping installation was integrated into an existing automated non-invasive high-throughput shoot phenotyping platform. Thus, the amended platform is now capable of conducting high-throughput phenotyping at the whole-plant level, and it was used to assess the vegetative root and shoot growth dynamics of five maize inbred lines and four hybrids thereof, as well as the responses of five inbred lines to progressive drought stress. The results showed that hybrid vigour (heterosis) occurred simultaneously in roots and shoots and was detectable as early as 4 days after transplanting (4 DAT; i.e., 8 days after seed imbibition) for estimated plant height (EPH), total root length (TRL), and total root volume (TRV). On the other hand, growth dynamics responses to progressive drought were different in roots and shoots. While TRV was significantly reduced 10 days after the onset of the water deficit treatment, the estimated shoot biovolume was significantly reduced about 6 days later, and EPH showed a significant decrease even 2 days later (8 days later than TRV) compared with the control treatment. In contrast to TRV, TRL initially increased in the water deficit period and decreased much later (not earlier than 16 days after the start of the water deficit treatment) compared with the well-watered plants. This may indicate an initial response of the plants to water deficit by forming longer but thinner roots before growth was inhibited by the overall water deficit. The magnitude and the dynamics of the responses were genotype-dependent, as well as under the influence of the water consumption, which was related to plant size.

Awn Image Analysis and Phenotyping Using BarbNet

Consideration of the properties of awns is important for the phenotypic description of grain crops. Awns have a number of important functions in grasses, including assimilation, mechanical protection, and seed dispersal and burial. An important feature of the awn is the presence or absence of barbs-tiny hook-like single-celled trichomes on the outer awn surface that can be visualized using microscopic imaging. There are, however, no suitable software tools for the automated analysis of these small, semi-transparent structures in a high-throughput manner. Furthermore, automated analysis of barbs using conventional methods of pattern detection and segmentation is hampered by high variability of their optical appearance including size, shape, and surface density. In this work, we present a software tool for automated detection and phenotyping of barbs in microscopic images of awns, which is based on a dedicated deep learning model (BarbNet). Our experimental results show that BarbNet is capable of detecting barb structures in different awn phenotypes with an average accuracy of 90%. Furthermore, we demonstrate that phenotypic traits derived from BarbNet-segmented images enable a quite robust categorization of 4 contrasting awn phenotypes with an accuracy of >85%. Based on the promising results of this work, we see that the proposed model has potential applications in the automation of barley awns sorting for plant developmental analysis.

Comparative metabolomics of root-tips reveals distinct metabolic pathways conferring drought tolerance in contrasting genotypes of rice

Background

The mechanisms underlying rice root responses to drought during the early developmental stages are yet unknown.

Results

This study aimed to determine metabolic differences in IR64, a shallow-rooting, drought-susceptible genotype, and Azucena, a drought-tolerant and deep-rooting genotype under drought stress. The morphological evaluation revealed that Azucena might evade water stress by increasing the lateral root system growth, the root surface area, and length to access water. At the same time, IR64 may rely mainly on cell wall thickening to tolerate stress. Furthermore, significant differences were observed in 49 metabolites in IR64 and 80 metabolites in Azucena, for which most metabolites were implicated in secondary metabolism, amino acid metabolism, nucleotide acid metabolism and sugar and sugar alcohol metabolism. Among these metabolites, a significant positive correlation was found between allantoin, galactaric acid, gluconic acid, glucose, and drought tolerance. These metabolites may serve as markers of drought tolerance in genotype screening programs. Based on corresponding biological pathways analysis of the differentially abundant metabolites (DAMs), biosynthesis of alkaloid-derivatives of the shikimate pathway, fatty acid biosynthesis, purine metabolism, TCA cycle and amino acid biosynthesis were the most statistically enriched biological pathway in Azucena in drought response. However, in IR64, the differentially abundant metabolites of starch and sucrose metabolism were the most statistically enriched biological pathways.

Conclusion

Metabolic marker candidates for drought tolerance were identified in both genotypes. Thus, these markers that were experimentally determined in distinct metabolic pathways can be used for the development or selection of drought-tolerant rice genotypes.

Brassica napus Roots Use Different Strategies to Respond to Warm Temperatures

Elevated growth temperatures are negatively affecting crop productivity by increasing yield losses. The modulation of root traits associated with improved response to rising temperatures is a promising approach to generate new varieties better suited to face the environmental constraints caused by climate change. In this study, we identified several Brassica napus root traits altered in response to warm ambient temperatures. Different combinations of changes in specific root traits result in an extended and deeper root system. This overall root growth expansion facilitates root response by maximizing root-soil surface interaction and increasing roots' ability to explore extended soil areas. We associated these traits with coordinated cellular events, including changes in cell division and elongation rates that drive root growth increases triggered by warm temperatures. Comparative transcriptomic analysis revealed the main genetic determinants of these root system architecture (RSA) changes and uncovered the necessity of a tight regulation of the heat-shock stress response to adjusting root growth to warm temperatures. Our work provides a phenotypic, cellular, and genetic framework of root response to warming temperatures that will help to harness root response mechanisms for crop yield improvement under the future climatic scenario.

Integrating advancements in root phenotyping and genome-wide association studies to open the root genetics gateway

Plant adaptation to challenging environmental conditions around the world has made root growth and development an important research area for plant breeders and scientists. Targeted manipulation of root system architecture (RSA) to increase water and nutrient use efficiency can minimize the adverse effects of climate change on crop production. However, phenotyping of RSA is a major bottleneck since the roots are hidden in the soil. Recently the development of 2- and 3D root imaging techniques combined with the genome-wide association studies (GWASs) have opened up new research tools to identify the genetic basis of RSA. These approaches provide a comprehensive understanding of the RSA, by accelerating the identification and characterization of genes involved in root growth and development. This review summarizes the latest developments in phenotyping techniques and GWAS for RSA, which are used to map important genes regulating various aspects of RSA under varying environmental conditions. Furthermore, we discussed about the state-of-the-art image analysis tools integrated with various phenotyping platforms for investigating and quantifying root traits with the highest phenotypic plasticity in both artificial and natural environments which were used for large scale association mapping studies, leading to the identification of RSA phenotypes and their underlying genetics with the greatest potential for RSA improvement. In addition, challenges in root phenotyping and GWAS are also highlighted, along with future research directions employing machine learning and pan-genomics approaches.

RootPainter: deep learning segmentation of biological images with corrective annotation

Convolutional neural networks (CNNs) are a powerful tool for plant image analysis, but challenges remain in making them more accessible to researchers without a machine-learning background. We present RootPainter, an open-source graphical user interface based software tool for the rapid training of deep neural networks for use in biological image analysis. We evaluate RootPainter by training models for root length extraction from chicory (Cichorium intybus L.) roots in soil, biopore counting, and root nodule counting. We also compare dense annotations with corrective ones that are added during the training process based on the weaknesses of the current model. Five out of six times the models trained using RootPainter with corrective annotations created within 2 h produced measurements strongly correlating with manual measurements. Model accuracy had a significant correlation with annotation duration, indicating further improvements could be obtained with extended annotation. Our results show that a deep-learning model can be trained to a high accuracy for the three respective datasets of varying target objects, background, and image quality with < 2 h of annotation time. They indicate that, when using RootPainter, for many datasets it is possible to annotate, train, and complete data processing within 1 d.

Comprehensive insights in thallium ecophysiology in the hyperaccumulator Biscutella laevigata

Biscutella laevigata is the strongest known thallium (Tl) hyperaccumulator plant species. However, little is known about the ecophysiological processes leading to root uptake and translocation of Tl in this species, and the interactions between Tl and its chemical analogue potassium (K). Biscutella laevigata was subjected to hydroponics experimentation in which it was exposed to Tl and K, and it was investigated in a rhizobox experiment. Laboratory-based micro-X-ray fluorescence spectroscopy (μ-XRF) was used to reveal the Tl distribution in the roots and leaves, while synchrotron-based μ-XRF was utilised to reveal elemental distribution in the seed. The results show that in the seed Tl was mainly localised in the endosperm and cotyledons. In mature plants, Tl was highest in the intermediate leaves (16,100 μg g^-1), while it was one order of magnitude lower in the stem and roots. Potassium did not inhibit or enhance Tl uptake in B.laevigata. At the organ level, Tl was localised in the blade and margins of the leaves. Roots foraged for Tl and cycled Tl across roots growing in the control soils. Biscutella laevigata has ostensibly evolved specialised mechanisms to tolerate high Tl concentrations in its shoots. The lack of interactions and competition between Tl and K suggests that it is unlikely that Tl is taken up via K channels, but high affinity Tl transporters remain to be identified in this species. Thallium is not only highly toxic but also a valuable metal and Tl phytoextraction using B. laevigata should be explored.

Improving the efficiency of plant root system phenotyping through digitization and automation

Root system architecture (RSA) determines unevenly distributed water and nutrient availability in soil. Genetic improvement of RSA, therefore, is related to crop production. However, RSA phenotyping has been carried out less frequently than above-ground phenotyping because measuring roots in the soil is difficult and labor intensive. Recent advancements have led to the digitalization of plant measurements; this digital phenotyping has been widely used for measurements of both above-ground and RSA traits. Digital phenotyping for RSA is slower and more difficult than for above-ground traits because the roots are hidden underground. In this review, we summarized recent trends in digital phenotyping for RSA traits. We classified the sample types into three categories: soil block containing roots, section of soil block, and root sample. Examples of the use of digital phenotyping are presented for each category. We also discussed room for improvement in digital phenotyping in each category.

RhizoVision Explorer: open-source software for root image analysis and measurement standardization

Roots are central to the function of natural and agricultural ecosystems by driving plant acquisition of soil resources and influencing the carbon cycle. Root characteristics like length, diameter and volume are critical to measure to understand plant and soil functions. RhizoVision Explorer is an open-source software designed to enable researchers interested in roots by providing an easy-to-use interface, fast image processing and reliable measurements. The default broken roots mode is intended for roots sampled from pots and soil cores, washed and typically scanned on a flatbed scanner, and provides measurements like length, diameter and volume. The optional whole root mode for complete root systems or root crowns provides additional measurements such as angles, root depth and convex hull. Both modes support providing measurements grouped by defined diameter ranges, the inclusion of multiple regions of interest and batch analysis. RhizoVision Explorer was successfully validated against ground truth data using a new copper wire image set. In comparison, the current reference software, the commercial WinRhizo™, drastically underestimated volume when wires of different diameters were in the same image. Additionally, measurements were compared with WinRhizo™ and IJ_Rhizo using a simulated root image set, showing general agreement in software measurements, except for root volume. Finally, scanned root image sets acquired in different labs for the crop, herbaceous and tree species were used to compare results from RhizoVision Explorer with WinRhizo™. The two software showed general agreement, except that WinRhizo™ substantially underestimated root volume relative to RhizoVision Explorer. In the current context of rapidly growing interest in root science, RhizoVision Explorer intends to become a reference software, improve the overall accuracy and replicability of root trait measurements and provide a foundation for collaborative improvement and reliable access to all.

Improvement of Phosphorus Use Efficiency in Rice by Adopting Image-Based Phenotyping and Tolerant Indices

Phosphorus is one of the second most important nutrients for plant growth and development, and its importance has been realised from its role in various chains of reactions leading to better crop dynamics accompanied by optimum yield. However, the injudicious use of phosphorus (P) and non-renewability across the globe severely limit the agricultural production of crops, such as rice. The development of P-efficient cultivar can be achieved by screening genotypes either by destructive or non-destructive approaches. Exploring image-based phenotyping (shoot and root) and tolerant indices in conjunction under low P conditions was the first report, the epicentre of this study. Eighteen genotypes were selected for hydroponic study from the soil-based screening of 68 genotypes to identify the traits through non-destructive (geometric traits by imaging) and destructive (morphology and physiology) techniques. Geometric traits such as minimum enclosing circle, convex hull, and calliper length show promising responses, in addition to morphological and physiological traits. In 28-day-old seedlings, leaves positioned from third to fifth played a crucial role in P mobilisation to different plant parts and maintained plant architecture under P deficient conditions. Besides, a reduction in leaf angle adjustment due to a decline in leaf biomass was observed. Concomitantly, these geometric traits facilitate the evaluation of low P-tolerant rice cultivars at an earlier stage, accompanying several stress indices. Out of which, Mean Productivity Index, Mean Relative Performance, and Relative Efficiency index utilising image-based traits displayed better responses in identifying tolerant genotypes under low P conditions. This study signifies the importance of image-based phenotyping techniques to identify potential donors and improve P use efficiency in modern rice breeding programs.

Fully-automated root image analysis (faRIA)

High-throughput root phenotyping in the soil became an indispensable quantitative tool for the assessment of effects of climatic factors and molecular perturbation on plant root morphology, development and function. To efficiently analyse a large amount of structurally complex soil-root images advanced methods for automated image segmentation are required. Due to often unavoidable overlap between the intensity of fore- and background regions simple thresholding methods are, generally, not suitable for the segmentation of root regions. Higher-level cognitive models such as convolutional neural networks (CNN) provide capabilities for segmenting roots from heterogeneous and noisy background structures, however, they require a representative set of manually segmented (ground truth) images. Here, we present a GUI-based tool for fully automated quantitative analysis of root images using a pre-trained CNN model, which relies on an extension of the U-Net architecture. The developed CNN framework was designed to efficiently segment root structures of different size, shape and optical contrast using low budget hardware systems. The CNN model was trained on a set of 6465 masks derived from 182 manually segmented near-infrared (NIR) maize root images. Our experimental results show that the proposed approach achieves a Dice coefficient of 0.87 and outperforms existing tools (e.g., SegRoot) with Dice coefficient of 0.67 by application not only to NIR but also to other imaging modalities and plant species such as barley and arabidopsis soil-root images from LED-rhizotron and UV imaging systems, respectively. In summary, the developed software framework enables users to efficiently analyse soil-root images in an automated manner (i.e. without manual interaction with data and/or parameter tuning) providing quantitative plant scientists with a powerful analytical tool.

Root responses to localised soil arsenic enrichment in the fern Pityrogramma calomelanos var. austroamericana grown in rhizoboxes

The terrestrial fern Pityrogramma calomelanos, a cosmopolitan tropical species, is one of the strongest known arsenic (As) hyperaccumulator plants. This study aimed to determine whether P. calomelanos preferentially forages for arsenite (As³⁺) or arsenate (As⁵⁺) in As-contaminated soils, and whether a positive root response to As enhances accumulation in P. calomelanos. Therefore, an experiment using rhizoboxes divided in two halves were constructed with a control soil (C) and As³⁺ or As⁵⁺ dosed soil at either 50 and 100 μg g^-1 As. Micro-X-ray Fluorescence elemental mapping (μXRF) was employed to analyze the distribution of As in roots and fronds, and Scanning Electron Microscopy with Energy Dispersive Spectroscopy (SEM-EDS) was used to determine As distribution in the reproductive tissues of P. calomelanos. The results showed that Pityrogramma roots do not specifically forage for As-contaminated soil; the area based on pixel counts was similar across all the treatments with no statistical differences. However, frond biomass was slightly higher in the treatments C ǀ As³⁺ and C ǀ As⁵⁺, and the highest accumulation of As in fronds was in the As⁵⁺ ǀ As³⁺ (100 μg g^-1) treatment, with 3418 and 2370 μg g^-1 in old and young fronds respectively. Arsenic cycling across the roots was observed by the μXRF mapping; in C ǀ As⁵⁺ (100) the As was higher and evenly distributed in both sections, whilst in C ǀ As³⁺ (50), the As was higher in the As³⁺ side. The μXRF mapping showed a broader As distribution in older fronds, where As was highest in the rachis and extended into the pinnule through the midrib. Pityrogramma calomelanos does not specifically root forage for As-enriched zones in the soil and grows healthily without signs of toxicity at lower (50 μg g^-1) and higher (100 μg g^-1) concentrations of As³⁺ and As⁵⁺ in the soil.

Detection and Classification of Bearing Surface Defects Based on Machine Vision

Surface defects on bearings can directly affect the service life and reduce the performance of equipment. At present, the detection of bearing surface defects is mostly done manually, which is labor-intensive and results in poor stability. To improve the inspection speed and the defect recognition rate, we proposed a bearing surface defect detection and classification method using machine vision technology. The method makes two main contributions. It proposes a local multi-neural network (Lc-MNN) image segmentation algorithm with the wavelet transform as the classification feature. The precision segmentation of the defect image is accomplished in three steps: wavelet feature extraction, Lc-MNN region division, and Lc-MNN classification. It also proposes a feature selection algorithm (SCV) that makes comprehensive use of scalar feature selection, correlation analysis, and vector feature selection to first remove similar features through correlation analysis, further screen the results with a scalar feature selection algorithm, and finally select the classification features using a feature vector selection algorithm. Using 600 test samples with three types of defect in the experiment, an identification rate of 99.5% was achieved without the need for large-scale calculation. The comparison tests indicated that the proposed method can achieve efficient feature selection and defect classification.

Effects of plant species and traits on metal treatment and phytoextraction in stormwater bioretention

To study effects of plant species selection on total and dissolved metal treatment performance of bioretention systems (BRS), 12 sets of columns were prepared, each planted with one of 12 species that are either widely used in BRS or have potentially important traits for metal removal (ability to hyperaccumulate metals, C4 photosynthesis, or ability to form mycorrhiza). Artificial stormwater was applied to half of the columns during all of a 31-week test period, while treatment of the others included a 5-week long dry period to test interactive effects of drying and plant traits on BRS metal treatment in more realistic alternating wet and dry conditions. Concentrations of metals (dissolved and total) in the effluent significantly differed between most columns with different plants, and the differences in concentrations of dissolved metals after the dry period were particularly important. Mean dissolved Cd concentrations exceeded Swedish reference values in effluents from BRS with two of the plant species, while mean dissolved Zn concentrations exceeded them in effluents from BRS with three of the species (and non-vegetated controls). Dissolved Cu leaching was observed in effluents from BRS with five of the plant species after the dry period, and mean concentrations exceeded Swedish reference values in effluents from all the BRS (including the constantly watered systems). Some support in terms of metal concentrations in shoots and shoot/soil ratios was obtained for using hyperaccumulators in BRS to remove metals from filter material. For example, Armeria maritima (a hyperaccumulator with the lowest shoot biomass) and Miscanthus sinsenis (a C4 plant with the highest biomass production) took up similar amounts of metals despite large differences in biomass. However, no significant correlations between effluent metal concentrations and plants' metal uptake were found, possibly because of the short duration of the experiment. The results indicate that root biomass affected effluent metal concentrations more strongly. Root biomass was often positively correlated with total and (particularly) dissolved effluent metal concentrations. Further experiments with different soil metal concentrations, organic matter analyses and stronger focus on root characteristics are recommended, including additional tests of effects of hyperaccumulators and mycorrhiza on metal treatment and phytoextraction.

Image-Based, Organ-Level Plant Phenotyping for Wheat Improvement

Wheat was one of the first grain crops domesticated by humans and remains among the major contributors to the global calorie and protein budget. The rapidly expanding world population demands further enhancement of yield and performance of wheat. Phenotypic information has historically been instrumental in wheat breeding for improved traits. In the last two decades, a steadily growing collection of tools and imaging software have given us the ability to quantify shoot, root, and seed traits with progressively increasing accuracy and throughput. This review discusses challenges and advancements in image analysis platforms for wheat phenotyping at the organ level. Perspectives on how these collective phenotypes can inform basic research on understanding wheat physiology and breeding for wheat improvement are also provided.

Non-destructive phenotyping for early seedling vigor in direct-seeded rice

BACKGROUND

Early seedling vigor is an essential trait of direct-seeded rice. It helps the seedlings to compete with weeds for water and nutrient availability, and contributes to better seedling establishment during the initial phase of crop growth. Seedling vigor is a complex trait, and phenotyping by a destructive method limits the improvement of this trait through traditional breeding. Hence, a non-invasive, rapid, and precise image-based phenotyping technique is developed to increase the possibility to improve early seedling vigor through breeding in rice and other field crops.

RESULTS

To establish and assess the methodology using free-source software, early seedling vigor was estimated from images captured with a digital SLR camera in a non-destructive way. Here, the legitimacy and strength of the method have been proved through screening seven diverse rice cultivars varying for early seedling vigor. In the regression analysis, whole-plant area (WPA) estimated by destructive-flatbed scanner (WPAs) and non-destructive imaging (WPA_i) approaches was strongly related (R² > 83%) and suggested that WPA_i can be adapted in place of destructive methods to estimate seedling vigor. In addition, this study has identified a set of new geometric traits (convex hull and top view area) for screening breeding lines for early seedling vigor in rice, which decreased the time by 80% and halved the cost of labor in data observation.

CONCLUSIONS

The method demonstrated here is affordable and easy to establish as a phenotypic platform. It is suitable for most glasshouses/net houses for characterizing genotypes to understand the plasticity of shoots under a given environment at the seedling stage. The methodology explained in this experiment has been proven to be practical and suggested as a technique for researchers involved in direct-seeded rice. Consequently, it will help in the simultaneous screening of genotypes in large numbers, the identification of donors, and in gaining information on the genetic basis of the trait to design a breeding program for direct-seeded rice.