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

Premise of the study X-ray microcomputed tomography (microCT) can be used to measure 3D leaf internal anatomy, providing a holistic view of tissue organisation. Previously, the substantial time needed for segmenting multiple tissues limited this technique to small datasets, restricting its utility for phenotyping experiments and limiting our confidence in the conclusion 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 which dramatically reduces the time required to process single leaf microCT scans into detailed segmentations. By training the model on each scan using 6 hand segmented image slices out of >1500 in the full leaf scan, it achieves >90% accuracy in background and tissue segmentation. Conclusion Overall, this 3D segmentation and quantification pipeline can reduce one of the major barriers to using microCT imaging in high-throughput plant phenotyping.

[1]  Thomas Brox,et al.  3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation , 2016, MICCAI.

[2]  Brendan Choat,et al.  Embracing 3D Complexity in Leaf Carbon-Water Exchange. , 2019, Trends in plant science.

[3]  S. Weiner,et al.  A 3D study of the relationship between leaf vein structure and mechanical function. , 2019, Acta biomaterialia.

[4]  A. McElrone,et al.  Maximum CO2 diffusion inside leaves is limited by the scaling of cell size and genome size , 2020, bioRxiv.

[5]  H. Sebastian Seung,et al.  Trainable Weka Segmentation: a machine learning tool for microscopy pixel classification , 2017, Bioinform..

[6]  S. Mooney,et al.  Mesophyll porosity is modulated by the presence of functional stomata , 2019, Nature Communications.

[7]  Emmanuelle Gouillart,et al.  scikit-image: image processing in Python , 2014, PeerJ.

[8]  M. Barbour,et al.  Cell and chloroplast anatomical features are poorly estimated from two-dimensional cross sections. , 2019, The New phytologist.

[9]  L. Sack,et al.  Outside-Xylem Vulnerability, Not Xylem Embolism, Controls Leaf Hydraulic Decline during Dehydration1[CC-BY] , 2017, Plant Physiology.

[10]  A. McElrone,et al.  Beyond Porosity: 3D Leaf Intercellular Airspace Traits That Impact Mesophyll Conductance1[OPEN] , 2018, Plant Physiology.

[11]  P. Verboven,et al.  Three-dimensional microscale modelling of CO2 transport and light propagation in tomato leaves enlightens photosynthesis. , 2016, Plant, cell & environment.

[12]  Gaël Varoquaux,et al.  Scikit-learn: Machine Learning in Python , 2011, J. Mach. Learn. Res..

[13]  Kevin W Eliceiri,et al.  NIH Image to ImageJ: 25 years of image analysis , 2012, Nature Methods.

[14]  Lisa Axe,et al.  Developments in synchrotron x-ray computed microtomography at the National Synchrotron Light Source , 1999, Optics & Photonics.

[15]  A. McElrone,et al.  The bias of a two-dimensional view: comparing two-dimensional and three-dimensional mesophyll surface area estimates using noninvasive imaging. , 2017, The New phytologist.

[16]  Travis E. Oliphant,et al.  Guide to NumPy , 2015 .

[17]  S. Wilkins,et al.  Simultaneous phase and amplitude extraction from a single defocused image of a homogeneous object , 2002, Journal of microscopy.

[18]  Fred A. Hamprecht,et al.  ilastik: interactive machine learning for (bio)image analysis , 2019, Nature Methods.

[19]  S. Mooney,et al.  Cell density and airspace patterning in the leaf can be manipulated to increase leaf photosynthetic capacity , 2017, The Plant journal : for cell and molecular biology.