Sample preparation


We aimed to create a sample preparation protocol that was cost effective and produced high quality plant sections. All plants included in this study exhibited transversely isotropic morphologies with vascular bundle fibers oriented in the longitudinal direction (e.g., see scanning electron microscopy [22] and computed tomography [12] images). The composite/transversely isotropic nature of plant stalks makes them especially difficult to cut without damaging the cross-sectional morphology of interest. Lignified vascular bundles are particularly difficult to cut without tearing or crushing of adjacent tissue structures. Common problems encountered when trying to section plant stalks include: tearing of vascular bundles, crushing of adjacent tissue structures, oblique faces of plant sections, and uneven/rough imaging surface. Table 1 shows the tools and blade types that were investigated in this study. Each of these tools/blade types was evaluated based on the cost of the tool and the quality of the resulting plant section.

Table 1 Evaluation of cutting machines for plant sectioning

The trim saw (Hi-Tech Diamond 6″ Trim Saw) [35] was set up with a Hi-tech Diamond Silver Thin Notched Saw Blade, which rotated between 800 and 3400 rpm, a saw fence and a saw vise [35]. Sections of maize, sorghum and poison hemlock plants were cut with the smooth, unpainted, and toothless saw blade at speeds typically between 1200 and 2000 rpm. The plant was held firmly with the saw vise and pushed through the rotating blade to cut sections. Sections were cut to required thickness by adjusting the saw fence accordingly. To avoid burnt section surfaces, it was essential to ensure pushing the samples through the rotating blade in a quick manner.

The water jet cutter (Omax Model 55100) [36] was operated on the fiberglass cutting setting. Sections produced with the water jet cutter were unburnt and undamaged but required relatively long time periods to cut and it created an uneven/oblique imaging surface. Sections with uneven surfaces subsequently resulted in poor image quality and unacceptable feature extraction results.

The angle grinder (Metabo WEV15-125) [37] and chop saw (Makita 2414DB Cut-Off Saw) [38] were mounted on a pivoted stand that had a built-in clamp to hold plant stalks firmly while cutting. The angle grinder was operated at a blade speed of 3000 rpm. The rotating blade was pushed down to cut through the clamped plant stalk. Plant stalk sections were cut with the angle grinder using three blade types:Piranha Tile Premium Diamond Blade (Tile blade) [39], T1 Premium Thin Cut-off Wheel (Fine blade) [40] and Resin-bonded Aluminum Oxide Flap Disc (Coarse blade) [41]. The tile blade produced the best results with the angle grinder as it produced unburnt sections and prevented ripped rinds. However, the tile blade pulled the vascular bundles embedded in the pith which caused poor image quality. The chop saw was set up with a 14″ abrasive cut off wheel which rotated at 3800 rpm blade speed. The chop saw produced damaged and unacceptable plant sections.

Sections of wheat and Arabidopsis plant stalks were cut using a razor blade. However, the wheat and Arabidopsis samples had to be hydrated before being sectioned by immersing them in distilled water. This reduced friction and enabled easier cutting of thin sections. The plant was held firmly with the thumb and index fingers in one hand and cut with the other. To avoid cutting of oblique sections the blade was held perpendicular to the sample and care was taken to ensure the blade was not bent while cutting. It was important to avoid any sawing motions while cutting sections and instead choping down in one single smooth motion. This prevented ripped rind and sample damage.


Plant sections were stained to improve contrast between plant structures thereby enhancing image analysis and feature extraction. Both simple and differential staining techniques were investigated. These included Alcian Blue, Safranin O, Toluidine Blue, and Alcian Blue-Safranin O staining techniques. The staining protocol for all sections included hydrating sections in distilled water, staining sections by submersion in a solution, rinsing in distilled water to remove excess staining solution followed by dehydrating the section in alcohol.

Differential staining of sections was achieved by following the Alcian Blue-Safranin O staining sequence (Fig. 10) [27].

Fig. 10

Alcian Blue-Safranin O staining sequence


We sought to employ an imaging setup with a wide field of view to enable the entire cross-section to be captured in a single picture. Adjustable magnification was desired to enable imaging of plant sections with different diameters. An AmScope LED Trinocular Zoom Stereo Microscope along with its 0.5× and 2.0× Barlow lenses and an 18MP digital camera [28] were selected as the best imaging option. The microscope had an objective lens with 0.7 to 4.5 magnification power when used along with its 0.5× and 2.0× Barlow lenses. The 0.5× lens enabled a large working distance for easy adjustment of sections and a wider field of view for the large cross-section of maize, sorghum, and poison hemlock sections. The 2× lens enabled extra zoom to capture smaller sized wheat and Arabidopsis sections.

The first step of the imaging process was to ensure the sharpest (i.e., focused) image of the stained section was viewed on the Amscope camera software interface (Toupview) [28] before capture. The next step was to calibrate the microscope to ensure accurate spatial conversion of section sizes. This was done using a scale rule to obtain the number of pixels equal to 1 mm. After calibration, section images were captured using the digital camera in combination with the AmScope software interface. Each image was saved in tagged image file (tif) format with a dimension of 4912 × 3684 pixels.

Feature extraction

Feature extraction from digitized plant images consists of four parts, namely: pre-processing, segmentation, smoothing, and quantification. The MATLAB (MATLAB R2019a) [29] computer programming software was used to develop an algorithm to implement each of these steps. The MATLAB code can be found in the provided Additional file 2. The MATLAB feature extraction algorithm was customized to work on both hollow and pith-filled plant types.

Pre-processing consisted of separating the pixels of digital images into groups to aid in distinguishing between different plant structures. To accomplish this the captured Red, Green and Blue (RGB) digital image was first converted to a grayscale image. This conversion reduced the red, green, and blue intensity channels to one intensity level per pixel. Pixels were then grouped according to their pixel intensity values using local adaptive thresholding [42] and Otsu thresholding [43] for pith-filled and hollow plant type respectively. Optimum threshold values were automatically calculated and applied to different regions of the image to aid in identifying plant structures. The image was then binarized to consist of either black or white pixels.

Segmentation involved identifying and separating different plant structures from the binary image. Maize and sorghum binarized sections were segmented into two different areas. The largest area (whole cross-section) in the binary image and the second largest area (the pith). Single material plant types were segmented into their hollow and whole cross-sectional areas. To ensure adequate identification and measurement of plant structures, image smoothing was applied to segmented areas.

Smoothing operations were implemented to remove unwanted details from segmented images. Unwanted details were identified as objects in images with very small area. The fill and erosion operations were used to smooth images [44]. Both processes required creating a new structuring element—an object with a specified radius or area and pixel, which is placed at the center of each object in the image to be smoothed [44]. Any image object smaller than the newly created object automatically picks up the pixel of the newly created object.

Quantification involved the extraction and measurement of identified plant structures. The rind boundary coordinates were used to quantify stalk diameters and rind thickness. The major and minor diameter endpoints were calculated using the rind outer boundary coordinates at 90°, 180°, 270° and 360o as shown in Fig. 1. The rind thickness was calculated as the mean of the shortest distance between each pair of the inner and outer rind boundary coordinates. These diameter and rind thickness measurements were validated in comparison to digital calipers measurements. The major and minor diameter were measured with digital calipers using the same orientation of the acquired image read into the image processing algorithm. The average rind thickness of each section was calculated by taking digital caliper measurements of the rind thickness at 90°, 180°, 270° and 360° as shown in Fig. 1.

In maize and sorghum sections, vascular bundles were quantified using a semi-automatic approach. First, the number of vascular bundles was automatically quantified. Then closely packed, collectively identified vascular bundles were deleted and individually re-selected. At this time any omitted bundles were also selected. The boundary of each vascular bundle was also extracted. These extracted rind and vascular bundle boundaries were imported into third party software for further analysis.

The image processing algorithm was developed to be compatible with microscope images of stained cross-sections of plant stalks. However, with minimal modification the algorithm can also work with other high-resolution and high-quality images. For example, the authors were able to use the algorithm to analyze and extract rind boundaries from a computed tomography image. However, when used in conjunction with low-resolution images obtained via a flatbed scanner, the algorithm was not as effective.

Exporting extracted plant geometry

Specimen specific cross-sectional geometries were extracted from digitized plant sections and imported into Abaqus software to create finite element models for analysis. In particular, the pixel location of geometric boundaries were converted to world coordinates in units of millimeters. This unit conversion was done using the MATLAB imREF2D function [29] and the calibration obtained during image acquisition. The MATLAB algorithm then created a Python script file that could be executed in Abaqus to create the specimen specific geometry. In particular, the MATLAB code printed the command statements required to create a model, specify a sheet size, sketch splines, and rename sketches in Abaqus. The MATLAB code also printed the rind and vascular bundle spline coordinates. Once the Python script executed in Abaqus it created a sketch as shown in Fig. 11.

Fig. 11

Imported maize stalk specimen-specific cross-sectional geometry

The specimen-specific geometries imported into Abaqus were used to create two-dimensional finite element models. These models were subjected to transverse compression analyses. Quantifying anisotropic material properties of plant stalks is challenging. However, it is common to use a transversely isotropic material assumption when modeling plant stems [4, 5, 11, 21, 23]. In this study the maize stalk rind and pith materials were modeled as transversely isotropic. The transverse Young’s modulus of the rind and pith materials were set as 850 MPa and 26 MPa, respectively. While the Poisson’s ratio was set as 0.25 [21]. As can be seen in Fig. 12, the geometries were meshed at a global seed size of 0.2 mm with four-node bilinear plane stress reduced integration quadrilateral elements with hourglass control. Two rigid body platens were used to compress the stalk model (see Fig. 12). The top platen was fixed in the rotational and horizontal direction and lowered in the vertical direction until it had compressed the stalk cross-section by 1 mm. The bottom platen was fixed in all degrees of freedom.

Fig. 12

Undeformed two-dimensional model of maize stalk in compression

Several three-dimensional finite element models were also created and subjected to bending loads. In particular, specimen-specific geometries of maize and wheat samples were used to create three-dimensional models. The three-dimensional geometry of the samples was created using specimen-specific cross-sectional geometry that was then extruded in the z-direction. The models were meshed with eight-node linear three-dimensional stress brick reduced integration hexahedral elements as shown in Fig. 13. One cross-sectional surface of the model was fixed in all degrees of freedom while a bending moment was applied to the opposite cross-sectional surface in the minor axis direction (see Fig. 14). Low bending moments of 1Nm were applied to the maize and wheat model. Documented material properties of maize and wheat were assigned to the models. For maize a Young’s modulus of 4.41GPa and 20 MPa were assigned to the rind and pith respectively with a Poisson’s ratio of 0.2 [11]. For wheat a Young’s modulus of value of 2.23GPa was assigned to the rind with a Poisson’s ratio of 0.413 [45].

Fig. 13

Undeformed three-dimensional model of maize (top) and wheat (bottom) internodes and their cross-sections

Fig. 14

Boundary conditions and loads applied to maize (top) and wheat (bottom) stalk internode models

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.


This article is autogenerated using RSS feeds and has not been created or edited by OA JF.

Click here for Source link (