Reading the leaves: A comparison of leaf rank and automated areole measurement for quantifying aspects of leaf venation

How dense can you be? New automatic measures of vein density in angiosperm leaves

Premise

Because of the trade-off between water loss and carbon dioxide assimilation, the conductivity of the transpiration path in a leaf is an important limit on photosynthesis. Closely packed veins correspond to short paths and high assimilation rates while widely spaced veins are associated with higher resistance to flow and lower maximum photosynthetic rates. Vein length per area (VLA) has become the standard metric for comparing leaves with different vein densities; its measurement typically utilizes digital image processing with varying amounts of human input.

Methods and Results

Here, we propose three new ways of measuring vein density using image analysis that improve on currently available procedures: (1) areole area distributions, (2) a sizing transform, and (3) a distance map. Each alternative has distinct practical, statistical, and biological limitations and advantages. In particular, we advocate the log-transformed modal distance map of a vein mask as an estimator to replace VLA as a standard metric for vein density.

Conclusions

These methods, for which open-source code appropriate for high-throughput automation is provided, improve on VLA by producing determinate measures of vein density as distributions rather than point estimates. Combined with advances in image quality and computational efficiency, these methods should help clarify the physiological and evolutionary significance of vein density.

An image dataset of cleared, x-rayed, and fossil leaves vetted to plant family for human and machine learning

Leaves are the most abundant and visible plant organ, both in the modern world and the fossil record. Identifying foliage to the correct plant family based on leaf architecture is a fundamental botanical skill that is also critical for isolated fossil leaves, which often, especially in the Cenozoic, represent extinct genera and species from extant families. Resources focused on leaf identification are remarkably scarce; however, the situation has improved due to the recent proliferation of digitized herbarium material, live-plant identification applications, and online collections of cleared and fossil leaf images. Nevertheless, the need remains for a specialized image dataset for comparative leaf architecture. We address this gap by assembling an open-access database of 30,252 images of vouchered leaf specimens vetted to family level, primarily of angiosperms, including 26,176 images of cleared and x-rayed leaves representing 354 families and 4,076 of fossil leaves from 48 families. The images maintain original resolution, have user-friendly filenames, and are vetted using APG and modern paleobotanical standards. The cleared and x-rayed leaves include the Jack A. Wolfe and Leo J. Hickey contributions to the National Cleared Leaf Collection and a collection of high-resolution scanned x-ray negatives, housed in the Division of Paleobotany, Department of Paleobiology, Smithsonian National Museum of Natural History, Washington D.C.; and the Daniel I. Axelrod Cleared Leaf Collection, housed at the University of California Museum of Paleontology, Berkeley. The fossil images include a sampling of Late Cretaceous to Eocene paleobotanical sites from the Western Hemisphere held at numerous institutions, especially from Florissant Fossil Beds National Monument (late Eocene, Colorado), as well as several other localities from the Late Cretaceous to Eocene of the Western USA and the early Paleogene of Colombia and southern Argentina. The dataset facilitates new research and education opportunities in paleobotany, comparative leaf architecture, systematics, and machine learning.

Automated and accurate segmentation of leaf venation networks via deep learning

Leaf vein network geometry can predict levels of resource transport, defence and mechanical support that operate at different spatial scales. However, it is challenging to quantify network architecture across scales due to the difficulties both in segmenting networks from images and in extracting multiscale statistics from subsequent network graph representations. Here we developed deep learning algorithms using convolutional neural networks (CNNs) to automatically segment leaf vein networks. Thirty-eight CNNs were trained on subsets of manually defined ground-truth regions from >700 leaves representing 50 southeast Asian plant families. Ensembles of six independently trained CNNs were used to segment networks from larger leaf regions (c. 100 mm² ). Segmented networks were analysed using hierarchical loop decomposition to extract a range of statistics describing scale transitions in vein and areole geometry. The CNN approach gave a precision-recall harmonic mean of 94.5% ± 6%, outperforming other current network extraction methods, and accurately described the widths, angles and connectivity of veins. Multiscale statistics then enabled the identification of previously undescribed variation in network architecture across species. We provide a LeafVeinCNN software package to enable multiscale quantification of leaf vein networks, facilitating the comparison across species and the exploration of the functional significance of different leaf vein architectures.

Super resolution for root imaging

PREMISE

High-resolution cameras are very helpful for plant phenotyping as their images enable tasks such as target vs. background discrimination and the measurement and analysis of fine above-ground plant attributes. However, the acquisition of high-resolution images of plant roots is more challenging than above-ground data collection. An effective super-resolution (SR) algorithm is therefore needed for overcoming the resolution limitations of sensors, reducing storage space requirements, and boosting the performance of subsequent analyses.

METHODS

We propose an SR framework for enhancing images of plant roots using convolutional neural networks. We compare three alternatives for training the SR model: (i) training with non-plant-root images, (ii) training with plant-root images, and (iii) pretraining the model with non-plant-root images and fine-tuning with plant-root images. The architectures of the SR models were based on two state-of-the-art deep learning approaches: a fast SR convolutional neural network and an SR generative adversarial network.

RESULTS

In our experiments, we observed that the SR models improved the quality of low-resolution images of plant roots in an unseen data set in terms of the signal-to-noise ratio. We used a collection of publicly available data sets to demonstrate that the SR models outperform the basic bicubic interpolation, even when trained with non-root data sets.

DISCUSSION

The incorporation of a deep learning-based SR model in the imaging process enhances the quality of low-resolution images of plant roots. We demonstrate that SR preprocessing boosts the performance of a machine learning system trained to separate plant roots from their background. Our segmentation experiments also show that high performance on this task can be achieved independently of the signal-to-noise ratio. We therefore conclude that the quality of the image enhancement depends on the desired application.

A microfluidic design to provide a stable and uniform in vitro microenvironment for cell culture inspired by the redundancy characteristic of leaf areoles

The leaf venation is considered to be an optimal transportation system with the mesophyll cells being divided by minor veins into small regions named areoles. The transpiration of water in different regions of a leaf fluctuates over time making the transportation of water in veins fluctuate as well. However, because of the existence of multiple paths provided by the leaf venation network and the pits on the walls of the vessels, the pressure field and nutrient concentration in the areoles that the mesophyll cells live in are almost uniform. Therefore, inspired by such structures, a microfluidic design of a novel cell culture chamber has been proposed to obtain a stable and uniform microenvironment. The device consists of a novel microchannel system imitating the vessels in the leaf venation to transport the culture medium, a cell culture chamber imitating the areole and microgaps imitating the pits. The effects of the areole and pit on flow fields in the cell culture chamber have been discussed. The results indicate that the bio-inspired microfluidic device is a robust platform to provide an in vivo like fluidic microenvironment.

Topological Phenotypes Constitute a New Dimension in the Phenotypic Space of Leaf Venation Networks

The leaves of angiosperms contain highly complex venation networks consisting of recursively nested, hierarchically organized loops. We describe a new phenotypic trait of reticulate vascular networks based on the topology of the nested loops. This phenotypic trait encodes information orthogonal to widely used geometric phenotypic traits, and thus constitutes a new dimension in the leaf venation phenotypic space. We apply our metric to a database of 186 leaves and leaflets representing 137 species, predominantly from the Burseraceae family, revealing diverse topological network traits even within this single family. We show that topological information significantly improves identification of leaves from fragments by calculating a “leaf venation fingerprint” from topology and geometry. Further, we present a phenomenological model suggesting that the topological traits can be explained by noise effects unique to specimen during development of each leaf which leave their imprint on the final network. This work opens the path to new quantitative identification techniques for leaves which go beyond simple geometric traits such as vein density and is directly applicable to other planar or sub-planar networks such as blood vessels in the brain.

Planar reticular networks are ubiquitous in nature and engineering, formed for instance by the arterial vasculature in the mammalian neocortex, urban street grids or the vascular network of plant leaves. We use a topological metric to characterize the way loops are nested in such networks and analyze a large database of 186 leaves and leaflets, revealing for the first time that the nesting of the networks’ cycles constitutes a distinct phenotypic trait orthogonal to previously used geometric features. Furthermore, we demonstrate that the information contained in the leaf topology can significantly improve specimen identification from fragments, and provide an empirical growth model that can explain much of the observed data. Our work can improve understanding of the functional significance of the various leaf vein architectures and their correlation with the environment. It can pave the way for similar analyses in diverse areas of research involving reticulate networks.

Phenoplant: a web resource for the exploration of large chlorophyll fluorescence image datasets

BACKGROUND

Image analysis is increasingly used in plant phenotyping. Among the various imaging techniques that can be used in plant phenotyping, chlorophyll fluorescence imaging allows imaging of the impact of biotic or abiotic stresses on leaves. Numerous chlorophyll fluorescence parameters may be measured or calculated, but only a few can produce a contrast in a given condition. Therefore, automated procedures that help screening chlorophyll fluorescence image datasets are needed, especially in the perspective of high-throughput plant phenotyping.

RESULTS

We developed an automatic procedure aiming at facilitating the identification of chlorophyll fluorescence parameters impacted on leaves by a stress. First, for each chlorophyll fluorescence parameter, the procedure provides an overview of the data by automatically creating contact sheets of images and/or histograms. Such contact sheets enable a fast comparison of the impact on leaves of various treatments, or of the contrast dynamics during the experiments. Second, based on the global intensity of each chlorophyll fluorescence parameter, the procedure automatically produces radial plots and box plots allowing the user to identify chlorophyll fluorescence parameters that discriminate between treatments. Moreover, basic statistical analysis is automatically generated. Third, for each chlorophyll fluorescence parameter the procedure automatically performs a clustering analysis based on the histograms. This analysis clusters images of plants according to their health status. We applied this procedure to monitor the impact of the inoculation of the root parasitic plant Phelipanche ramosa on Arabidopsis thaliana ecotypes Col-0 and Ler.

CONCLUSIONS

Using this automatic procedure, we identified eight chlorophyll fluorescence parameters discriminating between the two ecotypes of A. thaliana, and five impacted by the infection of Arabidopsis thaliana by P. ramosa. More generally, this procedure may help to identify chlorophyll fluorescence parameters impacted by various types of stresses. We implemented this procedure at http://www.phenoplant.org freely accessible to users of the plant phenotyping community.

Phenoplant: a web resource for the exploration of large chlorophyll fluorescence image datasets

BACKGROUND

Image analysis is increasingly used in plant phenotyping. Among the various imaging techniques that can be used in plant phenotyping, chlorophyll fluorescence imaging allows imaging of the impact of biotic or abiotic stresses on leaves. Numerous chlorophyll fluorescence parameters may be measured or calculated, but only a few can produce a contrast in a given condition. Therefore, automated procedures that help screening chlorophyll fluorescence image datasets are needed, especially in the perspective of high-throughput plant phenotyping.

RESULTS

We developed an automatic procedure aiming at facilitating the identification of chlorophyll fluorescence parameters impacted on leaves by a stress. First, for each chlorophyll fluorescence parameter, the procedure provides an overview of the data by automatically creating contact sheets of images and/or histograms. Such contact sheets enable a fast comparison of the impact on leaves of various treatments, or of the contrast dynamics during the experiments. Second, based on the global intensity of each chlorophyll fluorescence parameter, the procedure automatically produces radial plots and box plots allowing the user to identify chlorophyll fluorescence parameters that discriminate between treatments. Moreover, basic statistical analysis is automatically generated. Third, for each chlorophyll fluorescence parameter the procedure automatically performs a clustering analysis based on the histograms. This analysis clusters images of plants according to their health status. We applied this procedure to monitor the impact of the inoculation of the root parasitic plant Phelipanche ramosa on Arabidopsis thaliana ecotypes Col-0 and Ler.

CONCLUSIONS

Using this automatic procedure, we identified eight chlorophyll fluorescence parameters discriminating between the two ecotypes of A. thaliana, and five impacted by the infection of Arabidopsis thaliana by P. ramosa. More generally, this procedure may help to identify chlorophyll fluorescence parameters impacted by various types of stresses. We implemented this procedure at http://www.phenoplant.org freely accessible to users of the plant phenotyping community.

Bioinformatic and biometric methods in plant morphology1

Recent advances in microscopy, imaging, and data analyses have permitted both the greater application of quantitative methods and the collection of large data sets that can be used to investigate plant morphology. This special issue, the first for Applications in Plant Sciences, presents a collection of papers highlighting recent methods in the quantitative study of plant form. These emerging biometric and bioinformatic approaches to plant sciences are critical for better understanding how morphology relates to ecology, physiology, genotype, and evolutionary and phylogenetic history. From microscopic pollen grains and charcoal particles, to macroscopic leaves and whole root systems, the methods presented include automated classification and identification, geometric morphometrics, and skeleton networks, as well as tests of the limits of human assessment. All demonstrate a clear need for these computational and morphometric approaches in order to increase the consistency, objectivity, and throughput of plant morphological studies.