Dataset Information

Digitally deconstructing leaves in 3D using X-ray microcomputed tomography and machine learning.

ABSTRACT: Premise:X-ray microcomputed tomography (microCT) can be used to measure 3D leaf internal anatomy, providing a holistic view of tissue organization. Previously, the substantial time needed for segmenting multiple tissues limited this technique to small data sets, restricting its utility for phenotyping experiments and limiting our confidence in the inferences of these studies due to low replication numbers. Methods and Results:We present a Python codebase for random forest machine learning segmentation and 3D leaf anatomical trait quantification that dramatically reduces the time required to process single-leaf microCT scans into detailed segmentations. By training the model on each scan using six hand-segmented image slices out of >1500 in the full leaf scan, it achieves >90% accuracy in background and tissue segmentation. Conclusions:Overall, this 3D segmentation and quantification pipeline can reduce one of the major barriers to using microCT imaging in high-throughput plant phenotyping.

SUBMITTER: Theroux-Rancourt G

PROVIDER: S-EPMC7394714 | biostudies-literature | 2020 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Digitally deconstructing leaves in 3D using X-ray microcomputed tomography and machine learning.

Théroux-Rancourt Guillaume G Jenkins Matthew R MR Brodersen Craig R CR McElrone Andrew A Forrestel Elisabeth J EJ Earles J Mason JM

Applications in plant sciences 20200731 7

<h4>Premise</h4>X-ray microcomputed tomography (microCT) can be used to measure 3D leaf internal anatomy, providing a holistic view of tissue organization. Previously, the substantial time needed for segmenting multiple tissues limited this technique to small data sets, restricting its utility for phenotyping experiments and limiting our confidence in the inferences of these studies due to low replication numbers.<h4>Methods and results</h4>We present a Python codebase for random forest machine ...[more]

PMID: 32765979

Similar Datasets

Project description:BackgroundLeaf cellular architecture plays an important role in setting limits for carbon assimilation and, thus, photosynthetic performance. However, the low density, fine structure, and sensitivity to desiccation of plant tissue has presented challenges to its quantification. Classical methods of tissue fixation and embedding prior to 2D microscopy of sections is both laborious and susceptible to artefacts that can skew the values obtained. Here we report an image analysis pipeline that provides quantitative descriptors of plant leaf intercellular airspace using lab-based X-ray computed tomography (microCT). We demonstrate successful visualisation and quantification of differences in leaf intercellular airspace in 3D for a range of species (including both dicots and monocots) and provide a comparison with a standard 2D analysis of leaf sections.ResultsWe used the microCT image pipeline to obtain estimates of leaf porosity and mesophyll exposed surface area (Smes) for three dicot species (Arabidopsis, tomato and pea) and three monocot grasses (barley, oat and rice). The imaging pipeline consisted of (1) a masking operation to remove the background airspace surrounding the leaf, (2) segmentation by an automated threshold in ImageJ and then (3) quantification of the extracted pores using the ImageJ 'Analyze Particles' tool. Arabidopsis had the highest porosity and lowest Smes for the dicot species whereas barley had the highest porosity and the highest Smes for the grass species. Comparison of porosity and Smes estimates from 3D microCT analysis and 2D analysis of sections indicates that both methods provide a comparable estimate of porosity but the 2D method may underestimate Smes by almost 50%. A deeper study of porosity revealed similarities and differences in the asymmetric distribution of airspace between the species analysed.ConclusionsOur results demonstrate the utility of high resolution imaging of leaf intercellular airspace networks by lab-based microCT and provide quantitative data on descriptors of leaf cellular architecture. They indicate there is a range of porosity and Smes values in different species and that there is not a simple relationship between these parameters, suggesting the importance of cell size, shape and packing in the determination of cellular parameters proposed to influence leaf photosynthetic performance.

Project description:BackgroundWalnuts are grown worldwide in temperate areas and producers are facing an increasing demand. In a climate change context, the industry also needs cultivars that provide fruits of quality. This quality includes satisfactory filling ratio, thicker shell, ease of cracking, smooth shell and round-shaped walnut, and larger nut size. These desirable traits have been analysed so far using calipers or micrometers, but it takes a lot of time and requires the destruction of the sample. A challenge to take up is to develop an accurate, fast and non-destructive method for quality-related and morphometric trait measurements of walnuts, that are used to characterize new cultivars or collections in any germplasm management process.ResultsIn this study, we develop a method to measure different morphological traits on several walnuts simultaneously such as morphometric traits (nut length, nut face and profile diameters), traits that previously required opening the nut (shell thickness, kernel volume and filling kernel/nut ratio) and traits that previously were difficult to quantify (shell rugosity, nut sphericity, nut surface area and nut shape). These measurements were obtained from reconstructed 3D images acquired by X-ray computed tomography (CT). A workflow was created including several steps: noise elimination, walnut individualization, properties extraction and quantification of the different parts of the fruit. This method was applied to characterize 50 walnuts of a part of the INRAE walnut germplasm collection made of 161 unique accessions, obtained from the 2018 harvest. Our results indicate that 50 walnuts are sufficient to phenotype the fruit quality of one accession using X-ray CT and to find correlations between the morphometric traits. Our imaging workflow is suitable for any walnut size or shape and provides new and more accurate measurements.ConclusionsThe fast and accurate measurement of quantitative traits is of utmost importance to conduct quantitative genetic analyses or cultivar characterization. Our imaging workflow is well adapted for accurate phenotypic characterization of a various range of traits and could be easily applied to other important nut crops.

Dataset Information

Digitally deconstructing leaves in 3D using X-ray microcomputed tomography and machine learning.

Publications

Digitally deconstructing leaves in 3D using X-ray microcomputed tomography and machine learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets