A Biomimetic Study of Arctium minus (Burdock) Part II – Imaging Small (~mm) Hooks of Cellulose and Insect Chitin
As part of a project in producing a novel product from the burdock hook after biomimetic methodology developed by S N Gorb, this paper describes an investigation into the most appropriate method of shape acquisition for the purposes of reproduction and product development. This morphological study investigates confocal microscopy, 2-D digitising and the use of a microtome and digitising software. Small structures of cellulose and insect cuticle are imaged using confocal microscopy and the benefits and disadvantages of this approach are noted. A 3-D image of the burdock hook is produced from a 2-D digitised profile using SolidWorks 2004.
Keywords: Reverse engineering, shape acquisition, confocal microscopy, 2-D digitising, finite element analysis, rapid prototyping, Solid Works, cellulose, insect chitin, miniaturisation
1 Morphological Studies
This phase of the study concerns itself with recording shapes for the purposes of engineering analysis and reproduction.
1.1 Predictive and descriptive engineering
Predictive engineering is conventional engineering. A part is described with technical drawings and then analyzed to predict its behaviour and then constructed. Descriptive engineering is reverse engineering. An existing structure or mechanism is described and analyzed as it exists. Building restoration is a form of reverse engineering, particularly if the building is old and forgotten techniques are used to restore a building to its original condition. Engineered parts for which technical drawings have been lost or gone missing are reproduced using reverse engineering. Seeking to manufacture a product from a biological structure for which there never were any drawings, equally, is a form of reverse engineering.
Attempting to manufacture a structure that precisely mimics a biological structure is an attempt to unite both the prescriptive and descriptive, i.e. to use the prescriptive language of engineering to describe and analyze that which already exists, for the purposes of reproduction.
1.2 Shape Acquisition
Shape acquisition has a history in biological studies. From the first cave drawings man has endeavored to reproduce that which he observes in nature. Today, shape is used to provide clues as to internal composition of a biological structure when considered in the context of biomaterial strengths and behaviour and the use of the principle of shape optimization.
Dai, Gorb and Schwarz used methods of analyzing 2-D radii of curvature in insect tarsii to identify structural anomalies which superficially would seem to indicate zones of weakness or stress concentration but in reality identify zones of localized hardening/strengthening due to the presence of zinc or other trace minerals in the insect cuticle. When a structure does not break under loading when the shape of the structure would seem to indicate that it should, there is an indication that some material discontinuity is responsible.
Beraldin et al  in their paper on the virtual reality applications of scanning technology discuss the use of data transfer for layered manufacture and rapid prototyping. Confocal microscopy makes use of light intensities provided by fluorescing molecules to form images of minute structures and their internal components. Evans et al  used this physical phenomenon and technology, by casting the external morphologies of bat’s teeth to generate 3-D images of teeth to study wear patterns.
Finite element analysis comprises the precise division of a 3-D morphology into vertices and edges in order to compute stresses at a distance from an applied load in a structure. Surface modeling using constructs such as the Canny edge detection method to create order out of data clouds, transforming them into a triangulated form for the purposes of creating 3-D surfaces.
The above introduction prompted the following questions:
- Can a mesh formed from random points of light intensities be used to form a finite element analysis mesh?
- Can a mesh formed from random points of light intensities be used to form a mesh for conversion to .stl format and sent to a rapid prototyping device?
- Can small structures be imaged using a confocal microscope without the use of the casting methods of Evans et al?
2.1 Microscopy Techniques
The following techniques were the focus of preliminary investigation, through the lectures of Dr I Jones, then of the Neuroscience Department at the University of Bath :
- The principles of fluorescence microscopy
- Epi-fluorescence microscopy
- Confocal microscopy
- 2-photon microscopy
- Near field scanning optical microscopy
The fluorescence effect is produced by irradiating atoms with a high energy light source (laser) which causes excitation of orbiting electrons. These electrons jump “outwards” to high energy orbitals before returning to their normal state, releasing energy at a specific wavelength which is detected via an emission filter.
Confocal microscopy makes use of a laser light source whereas epi-fluorescence microscopy makes use of a normal bright light source and two filters, an excitation filter and an emission filter. Samples are viewed through an eye-piece.
The advantages of confocal microscopy using a laser light are:
- Reduced blurring
- Increased effective resolution
- Improved signal to noise ratio
- z-axis scanning
- depth perception
- magnification is electronically adjusted
- there is clear examination of thick specimens
The following procedure is described by Evans et al for the production of cubic voxels and virtual reality applications from his paper on the imaging of mammalian teeth . It essentially notes how to take a suitable image of a small object (~mm) for the purposes of digitizing, reproduction and study in virtual reality.
In capturing an image it must be born in mind that the goal was an accurate 3-D model for both virtual reality applications. It is important to set the slice thickness accordingly to arrive at an undistorted image i.e. cubic voxels. The paper by Evans et al details a method of taking a cast of a tooth which is more technically cumbersome than simply putting a microscope slide with specimen under the objective and so some of his paper is not relevant here (the details concerning the casting of the teeth).
Optical slices were taken through the x, y plane where each slice was square (e. g. 256 x 256) pixel 8-bit image at medium scanning speed.
Slices must be taken at the same distance as the interval between pixels to make cubic voxels.
Software such as Zeiss is used to generate a 3-D image from the stack of slices, where pixel intensity represents height and the z-height is found by comparing the intensities for each x, y point (in fact, a column of pixels all with co-ordinates (x, y)). In most of the tests run by Evans et al, the cubic voxels (and z-interval) were 7.8 mm long, generated in one of two methods:
- For the x 5mm lens – a 256 x 256 pixel image was scanned at zoom 1 (field of view (FOV) of 2 x 2 mm), or a 128 x 128 pixel image was scanned at zoom 2 (FOV 1 x 1 mm)
- For the x 10mm lens, a 128 x 128 pixel image was scanned at zoom 1 (FOV 1 x 1mm)
Therefore using a lens with a field of view (FOV) of 2 x 2 mm at a setting of zoom 2 reduces the field of view to 1 x 1.
Surface noise can affect the image and give a false indication of where the true surface lies. Evans et al did experiments with the x 5 and x 10 lens to see how best to obtain the most accurate surface image. They used two techniques to try to reduce surface noise; accumulation and averaging. Accumulation is to accumulate and average several images at each z height and then create an image from the accumulated image slices. On the microscope, for example, an “Accumulation 2” scan stands for the number of slices that are averaged (two).
The second method was to take the average of a number of reconstructed 3-D images of the same area. This was tested using a specially prepared and dimensionally precise standard glass specimen and comparing resultant images. The specimen was cubic and so without any undercuts but with a 45o fillet. Inner width was 1.3mm and outer width 1.7mm.
It was found that averaging produces better results than accumulation. Sanson et al used a resin casting of their teeth specimens which was coated with eosin, a fluorescent dye.
It was decided that small biological specimens could possibly be translucent enough to laser light such that it might not be necessary to use the casting method of Evans et al. Instead it was decided to attempt to image plain untreated specimens of cellulose and insect chitin.
A specimen burdock bract was mounted upon a “well” microscope slide in distilled water (it is a feature of both confocal and atomic force microscopy that specimens may be mounted without treatment) and placed under the objective of a confocal microscope. (Sincere thanks are due to Dr Ian Jones, post-doctoral researcher in Neuroscience in the Biology Department, University of Bath for his curiosity, assistance and instruction on operating the microscope.).
- A Zeiss Axiovert single photon confocal microscope (inverted microscope with the objectives underneath the platform).
- Zeiss LSM 510 module (laser scanning microscope) with 2 lasers
- 1 x Argon (488nm)
- 2 x HeNe (543nm & 633nm)
- all x 10, 40(oil), 63(oil) & 63(water)
- digital zoom up to x 200
- differential interference contrast.
The field of view: 1 x 1 mm.
Pinhole setting: 1 optical unit.
Scanning slice thickness: 19nm
- Well slides which are microscope slides with a bowl ground out in the centre to receive specimens that are not flat.
- Distilled water as a medium for slide mounting.
- It is important to get the hooked specimen in the right orientation on the slide to avoid displaying an undercut surface to the laser light.
- It is important to optimise the strength of the laser and reduce the required depth of penetration to prevent excess bleaching.
- The same specimen can be remounted a number of times in different orientations in the slide to fully expose the complete detail of the structure.
When suffused with the laser light at three different frequencies it was found that the burdock hook fluoresced well under the green laser light. Under the red and blue light the resulting image was less distinct but these colours worked well for the insect tarsii. The stacked image is then output to file and stored as a sequence of .tif files that are viewed in .avi format (see below for the full range of .tif images).
Stereogram images of the hook follow (Figure 1, Figure 2, Figure 3). The data from the confocal microscope is a sequence of image slices that are then automatically reassembled (stacked). Evidence of the stacking can be observed in the images from the stepped outline of each image. The glow that surrounds the stereogram images derives from the fact that this view of the hook is assembled using standard confocal software and the viewer is looking through preceding and following images which are a result of the perspective of looking at angled images. Only in a profile image does a stark outline of the hook show. There is an artefact on the microscope slide that shows to the side of the hook.
Figure 1:Stereogram 1 of the burdock hook specimen
Figure 2:Stereogram 2 of the burdock hook specimen
Figure 3:Stereogram 3 of the burdock hook specimen
(Dr I Jones, October 2002)
3.1.1 Burdock .tif images
The individual .tif images that make up the above stereogram are below (see Figure 8) imaged under the green light. The specimen was lying upon its side for the z-axis scan to minimise the number of scans required to scan the entire specimen, with the hook in profile to take into account undercut of the hook.
1. 2. 3.
4. 5. 6.
7. 8. 9.
Figure 4:1 – 20 The individual z-axis scan .tif files that make up the stereogram of the burdock hook (the scale bar defines 200 microns) (Dr I Jones October 2002).
Note that the images from the confocal microscope show some internal structure of the hook, particularly the cellulose microfibrils. These microfibrils are visible in the next experiment which fractures the hooks in a tensile tester. The hooks are made up of cellulose fibres bound together with hemi-cellulose to form microfibrils . The curves of the hook are smooth suggesting that the material is homogenous.
Figures 5 and 6 show the tarsii of two insects, a common grasshopper and a common bee, both composed of insect chitin. Tarsi and setae are clearly visible.
3.2 Grasshopper .tif images
1 2 3 4
5 6 7 8
Figure 5:1 – 30 The individual z-axis scan .tif files of the scan through the tarsus of a common grasshopper (the scale bar defines 200 microns) (Dr I Jones October 2002).
3.3 Bee .tif images
1 2 3 4
5 6 7 8
Figure 6:1 – 30 The individual z-axis scan .tif files of the scan through the tarsus of a bee (the scale bar defines 200 microns) (B Saunders October 2002).
Can a mesh formed from random points of light intensities be used to form a finite element analysis mesh? No, because finite element analysis requires discrete points to form a continuous scaffold for the purposes of calculation. Only carefully constructed shapes in finite element software can be meshed for the purposes of structural analysis. Edge detection and data clouds of light intensities do not yield the ordered data sets required to form a finite element scaffold.
Can a mesh formed from random points of light intensities be used to form a mesh for conversion to .stl format and sent to a rapid prototyping device? Yes, as is indicated by Beraldin et al above. There is software available on the market that allows for file conversion to .stl format required for a rapid prototyping device. There needs to be some consideration as to the purpose of the imaging. There are limitations to the capabilities of a rapid prototyping device which means that any reproduction would lose its scaling effects. Sanson et al  used confocal microscopy combined with taking casts of bat’s teeth to produce images of tooth wear patterns to be studied using virtual reality. Beraldin et al described a variation of the technique which could be applied to any stack of data points to produce a resin model from a rapid prototyping device or a model in virtual reality applying it to works of sculpture but the question needs to be asked: of what practical use would a large scale resin model of a biological structure be apart from fulfilling some educational role? The virtual reality applications particularly with respect to the medical field would seem more important.
Can small structures be imaged using a confocal microscope without the use of casting methods? Yes. The first method of shape recording selected was the use of confocal microscopy. Confocal microscopy has a particular attraction due to the fact that specimens need only be mounted in a distilled water solution and the examination is non-destructive. The ambition of this experiment was to seek a method of direct data transfer without recourse to curve-fitting for moving directly to a prototyping device.
Finite Element Analysis cannot be abandoned for the purposes of product design therefore confocal microscopy and the random mesh arising from its imaging loses its attraction. Instead a more ordered form of image acquisition is used. Why? Because the purpose is to manufacture a product out of a standard material and there is a need to predict behaviour.
Confocal microscopy is commonly used in the field of neuroscience. Its application to the field of biomimetic study could be controversial since it is expensive hardware and memory intensive. This notwithstanding, memory capacities are increasing and technology advancing at a speed which may make its use more frequent in the future.
In reality a confocal microscope and its output is a laser scanner like many others on the market, engineered for microscopic applications. The random nature of measurement of light intensities is suitable for conversion to .stl files and this was shown by Dr Dylan Evans who used an MRI scan, also an output of image sections, to construct a model of the human heart using a rapid prototyping device. But it is not suitable for the ordered nature of Finite Elelment Analysis.
There is not yet a unifying and truly descriptive and prescriptive method of shape acquisition. The compromise is creating a mesh for the purpose of surface modelling and a second mesh for the purpose of finite element analysis, each mesh deriving from a different data set. And because commercial design packages include .stl compatibility the following method was adopted.
4 Alternative Morphological Recording: Sectioning and 2-D digitizing
It is appropriate to consider an alternative means of collecting morphological data on the hooks. This comprises of using a microtome and a digitizer with a finite element package such as Solid Modeling. One begins by taking into account that one shall be using a Solid Modeling feature known as “loft”. This requires what is known as a “path” which is a digitized spine along which a number of cross-sections shall be distributed prior to rendering of the object in 3D. Digitizing the upper and lower profiles provides for a smoother model by providing two paths to support the intervening profiles. Thereafter, using the microtome to section at predefined and measured intervals along the profile, a number of cross-sections are taken perpendicular to the inner spline and each digitized. Each section is preserved so that the sectioned material can be inspected. This method would seem to be particularly applicable to the biomimetic study of insect tarsi.
The data is transferred to the Solid Modeling package to reconstruct the hook in 3 dimensions. The advantage to this form of morphological recording is that it is cheaper and mobile compared to a confocal microscope – the digitizer, microtome and software are relatively commonplace, the digitized sections are distinct and internal detail is revealed by the sectioning and this can be included in the reconstruction.
Alternatively a simple silhouette can be used, measured and reconstructed. It was assumed each profile was circular which is clearly not entirely true but is a close enough approximation for the purposes of illustration.
Figure 7:Figure 1 – Electron micrograph with grid superimposed. Each bar represents an interval of 100 microns.
Using Figure 7 it was possible to digitise two splines for the inner and outer profiles. Diameters were measured perpendicular to the inner spline and used to reconstruct the hook using the loft feature, the result of which is shown below.
Figure 8:3-D image of reconstructed burdock hook. For scale compare with scale bar of Figure 7.
The generation of a product from a biological structure combines prescriptive and descriptive processes. At present technology does not allow us to make the transition smoothly. Finite element analysis and graphical representation utilise the 3-D meshing of data points for differing purposes.
Small biological structures of cellulose and insect chitin are translucent to laser light and therefore can be scanned with non-destructive results. This may also be true of other biomaterials. Commercially available software can be utilised to convert the resultant dataclouds to 3-D models but this does not mean that the model is suitable for finite element analysis.
In terms of efficiency based upon cost and computational effectiveness including memory storage, confocal microscopy is more expensive than conventional methods of sectioning and digitising. However in terms of operator time it is by far the cheapest.
It could be possible to use the confocal microscope to perform data capture and export the product to a finite element package, thereafter to re-mesh to perform finite element analysis. This is not so simple as it might seem, however, since finite element analysis has its own demands in terms of discrete vertices and lengths that must be fulfilled via the act of drawing construction by the user. These needs are not fulfilled by appointing relatively random vertices of light intensities that arise through the act of laser scanning a biological structure.
It has been established in Part I that the burdock hook consists of a uniform biomaterial which is known to be anisotropic. It is sensitive to shear due to the bending moment of loading and resistant to flexing. It is formed to shear under heavy loading and to be receptive to many different substrates. The burdock hook, through its flattened bract support, has a single degree freedom which increases its propensity to attach.
 Dai Z, Gorb S N, Schwarz U,Roughness-dependent friction force of the tarsal claw system in the beetle Pachnoda marginata (Coleoptera, Scarabaeidae), Journal of Experimental Biology, 205, pp. 2479-2488, 2002
 Beraldin J. A., Blais F., Boulanger P., Cournoyer L., Domey J, El-Hakim S. F., Godin G., Rioux M., Taylor J., Real world modelling through high resolution digital 3D imaging of objects and structures. ISPRS Journal of Photogrammetry and Remote Sensing, 55, pp. 230-250, 2000
 Evans A. R, Harper I S, Sanson G D,Confocal imaging, visualisation and 3-D surface measurement of small mammalian teeth. Journal of Microscopy, 204, Pt 2 pp. 108-119 2001
 Jones I., Personal Communication, November 2002, Research Officer, Neuroscience, Department of Biology, University of Bath, Bath, UK
 Vincent J. F. V., Structural Biomaterials, The Macmillan Press, 1982
 Sellinger A, Weiss P. M., Nguyen A., Lu Y., Assink R. A., Gong W., Gong C., Brinker C. J., Continuous self-assembly of organic-inorganic nanocomposite coatings that mimic nacre. Nature, 394, pp. 256-260 (1998)
 Devlin R. M., Witham F. H., Plant Physiology, Fourth Ed., Devlin and Witham, PWS, 1983