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

Abstract

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.

Maximum CO2diffusion inside leaves is limited by the scaling of cell size and genome size

“…Images were reconstructed using TomoPy [ 54 ] for all ALS samples or using the in-house reconstruction platform for SLS or APS samples. Reconstructed scans were processed using published methods [ 32 , 55 ], and image stacks were cropped to remove tissue that was dehydrated, damaged or contained artefacts from the imaging or reconstruction steps. The final stacks contained approximately 500–2000 eight-bit grayscale images (downsampled from 16 or 32-bit images).…”

“…To extract surface area and volumes, mesophyll cells, airspace, vasculature (combined veins and bundle sheath) and background (including the epidermis) were segmented using published methods [ 32 , 55 ] and ImageJ [ 57 ]. Airspace volume ( V pores ), mesophyll cell volume ( V cells ), both summing up to the total mesophyll volume ( V mes ), vasculature volume ( V veins ) and the surface area exposed to the intercellular airspace (SA mes ) were then extracted using published methods [ 32 ] with the ImageJ plugin BoneJ [ 58 ], or using a custom Python program [ 55 ] ( ). SA mes / V mes is less sensitive to leaf thickness than the commonly measured S m , i.e.…”

Advances on the Visualization of the Internal Structures of the European Mistletoe: 3D Reconstruction Using Microtomography

“…The segmentation of the gray values of microCT data sets is increasingly being automated by algorithms due to the advancing technical possibilities ( Théroux-Rancourt et al, 2020 ). However, our study demonstrated that especially when several tissues with (almost) the same gray level are located close to each other, a in-depth biological knowledge and a considerable amount of time are required for a precise segmentation (or, in the best case, to verify the correctness of the segmentation after automated segmentation).…”

The three‐dimensional construction of leaves is coordinated with water use efficiency in conifers

“…The combination of the binary image derived from both reconstruction methods, along with the tissue boundaries, resulted in a composite image stack where each leaf tissue was classified. Leaf segmentation, which allowed us to automatically delimit different tissues across the full stack using a limited set of hand‐segmented composite slices was done using random‐forest classification (Théroux‐Rancourt et al ., 2020).…”