Use of X-ray computed microtomography for non-invasive determination of wood anatomical...

Journal of

Structural

Journal of Structural Biology 148 (2004) 11–21

Biology

www.elsevier.com/locate/yjsbi

Use of X-ray computed microtomography fornon-invasive determination of wood anatomical characteristics

Kathy Steppe,a,* Veerle Cnudde,b Catherine Girard,b,c Raoul Lemeur,a

Jean-Pierre Cnudde,c and Patric Jacobsb

a Laboratory of Plant Ecology, Ghent University, Belgiumb Laboratory of Sedimentary Geology and Engineering Geology, Ghent University, Belgium

c Laboratory of Applied Geology and Hydrogeology, Ghent University, Belgium

Received 23 January 2004, and in revised form 7 May 2004

Available online 9 June 2004

Abstract

Quantitative analysis of wood anatomical characteristics is usually performed using classical microtomy yielding optical mi-

crographs of stained thin sections. It is time-consuming to obtain high quality cross-sections from microtomy, and sections can be

damaged. This approach, therefore, is often impractical for those who need quick acquisition of quantitative data on vessel

characteristics in wood. This paper reports results of a novel approach using X-ray computed microtomography (microCT) for

non-invasive determination of wood anatomy. As a case study, stem wood samples of a 2-year-old beech (Fagus sylvatica L.) and

a 3-year-old oak (Quercus robur L.) tree were investigated with this technique, beech being a diffuse-porous and oak a ring-porous

tree species. MicroCT allowed non-invasive mapping of 2-D transverse cross-sections of both wood samples with micrometer

resolution. Self-developed software �lCTanalysis� was used for image processing of the 2-D cross-sections in order to automat-

ically determine the inner vessel diameters, the transverse cross-sectional surface area of the vessels, the vessel density and the

porosity with computer assistance. Performance of this new software was compared with manual analysis of the same micro-

graphs. The automatically obtained results showed no significant statistical differences compared to the manual measurements.

Visual inspection of the microCT slices revealed very good correspondence with the optical micrographs. Statistical analysis

confirmed this observation in a more quantitative way, and it was, therefore, argued that anatomical analysis of optical mi-

crographs can be readily substituted by automated use of microCT, and this without loss of accuracy. Furthermore, as an

additional application of microCT, the 3-D renderings of the internal microstructure of the xylem vessels for both the beech and

the oak sample could be reconstructed, clearly showing the complex nature of vessel networks. It can be concluded that the use of

microCT in wood science offers an interesting potential for all those who need quantitative data of wood anatomical charac-

teristics in either the 2-D or the 3-D space.

� 2004 Elsevier Inc. All rights reserved.

Keywords: Microtomography (microCT); X-ray; Microtomy; Micrograph; Light microscopy; Image processing; Double thresholding; 3-D image;

Fagus sylvatica L.; Quercus robur L.; Beech; Oak

1. Introduction

Many research topics need the quantitative analysis

of wood anatomical characteristics. For instance, in

palaeoecological research, the diameter of successivetracheid cells measured across growth rings may be used

to distinguish wood tissue produced by deciduous and

* Corresponding author. Fax: +32-9-224-44-10.

E-mail address: [email protected] (K. Steppe).

1047-8477/$ - see front matter � 2004 Elsevier Inc. All rights reserved.

doi:10.1016/j.jsb.2004.05.001

evergreen fossil coniferopsid species (Falcon-Lang,

2000). In tree physiological research, the specific hy-

draulic conductivity of the wood matrix in stems and

branches is often estimated theoretically from xylem

anatomical characteristics using the Hagen–Poiseuille�sequation (e.g., James et al., 2003). This specific hydraulic

conductivity of the stem and the branches may change

when living trees are subjected to water stress (Lovisolo

and Schubert, 1998), low-light conditions (Hoffmann

and Schweingruber, 2002; Schultz and Matthews, 1993)

mail to: [email protected]

12 K. Steppe et al. / Journal of Structural Biology 148 (2004) 11–21

or elevated CO2 concentrations (Atkinson and Taylor,1996; Gartner et al., 2003) as the vessel size and/or the

amount of vessels per unit xylem area may change

during these events. Quantification of wood anatomical

characteristics is required in all of these areas.

To date, quantification of the xylem anatomy has

mainly been done using classical microtomy; often in

combination with digital analysis of optical micrographs

obtained from stained thin sections. However, classicalcross-section microtomy is not always practical because

of difficulties related to the time and effort needed for

sample preparation, and to the probability of damage

incurred during sample preparation (Bentley et al.,

2002).

X-ray computed microtomography (microCT), be-

ing the high resolution variant of the medical CT

scanner, overcomes such limitations. A series of non-invasive views through the sample (i.e., radiographic

projections recorded from different viewing angles)

are reconstructed mathematically into transverse two-

dimensional (2-D) cross-sections with micrometer

resolution. Moreover, X-ray microCT also provides a

means to study the three-dimensional (3-D) internal

microstructure of the sample. In combination with

adequate software, automatic image processing of the2-D and the 3-D images can be performed in limited

time.

X-ray microCT has found its application in a

number of research disciplines, e.g., soil science

(Perret et al., 1999), geology (Bentz et al., 2000;

Cnudde and Jacobs, 2003; Van Geet et al., 2001),

hydrology (Wildenschild et al., 2002), plant physiol-

ogy (Fromm et al., 2001) and human physiology(Bentley et al., 2002). The purpose of this paper is to

demonstrate that microCT can also be successfully

applied within the scope of wood anatomical re-

search. It is tested whether transverse cross-sections

obtained with this novel technique can be used for

quantifying wood anatomical characteristics instead of

optical micrographs of stained thin sections. Self-de-

veloped software was thereby used to analyse themicroCT images in order to automatically determine

the inner vessel diameter, the transverse cross-sec-

tional surface area of the vessels, the vessel density

and the porosity. Particular attention was paid to the

correct performance and accuracy of this new soft-

ware. As a case study, samples of stem wood for a

young beech (Fagus sylvatica L.) and a young oak

(Quercus robur L.) tree were analysed for differencesin vessel anatomy. It is known that beech has a dif-

fuse-porous wood matrix, while oak has ring-porous

wood. For this purpose, the performance of X-ray

microCT was evaluated against use of optical micro-

graphs obtained by conventional microtomy. The

potential of microCT for 3-D renderings of xylem

microstructure was also evaluated.

2. Material and methods

2.1. Plant material

A 2-year-old beech tree (F. sylvatica L.) and a 3-year-

old oak tree (Q. robur L.) were used as test trees. They

were part of a study on the dynamics of sap flow of

young trees in relation to xylem characteristics (Steppe

and Lemeur, 2004). Stem wood was sampled when thetrees were about 1.30m high with a stem diameter at the

soil surface of 11.7 and 12.7mm, respectively. The trees

were grown in a container (0.4m diameter and 0.4m

height) filled with a silt loam soil. The test trees were

placed in the growth room of the Laboratory of Plant

Ecology (Ghent University, Belgium), where day and

night temperature was regulated to be 23 and 13 �C re-

spectively. The intensity of radiation from the fluores-cent lamps (�TL�D 80, Philips Lighting) was maintained

to a constant photon flux density of about 300 lmol

PAR m�2 s�1 during 14 h/day. The choice of the beech

and the oak tree as test trees was based on their differ-

ence in wood anatomy, as the xylem tissue of beech is

classified as diffuse-porous and oak xylem tissue as ring-

porous. After eight months of sap flow experiments in

the growth room, the stems of the young trees were cutinto pieces of about 30mm length. The cut stem

segments were softened and stored in a mixture of

glycerine–ethanol (50% v/v) until they were used for

either the novel X-ray microCT technique or the

classical microtomy.

2.2. X-ray computed microtomography (microCT)

Small wood cubes of 5� 5� 25mm (B�H�L) were

cut out of the stored stem segments for further analysis.

To obtain X-ray microtomographical images of the

wood samples a �SkyScan 1072� microtomograph was

used. This is a compact desktop system for X-ray mi-

croscopy and computed microtomography. It consists of

the combination of an X-ray shadow microscope sys-

tem, containing a microfocus sealed X-ray tube withhigh-voltage power supply, an object stage with preci-

sion manipulator, a 2-D X-ray CCD-camera connected

to the digital frame grabber and a computer with

tomographic reconstruction software. The X-ray

CCD-camera is based on a high-resolution (1024� 1024

pixels) cooled CCD-sensor with fibre optic coupling to

an X-ray scintillator.

During image acquisition, X-ray radiographs wererecorded at different angles during step-wise rotation

(steps of 0.45�) of the sample between 0� and 180�around the vertical axis. The X-ray source and the de-

tector remained in a fixed position. The samples were

scanned at a source voltage of 130 kV and a current of

76 lA. A random synchronous 10-pixel movement

in vertical direction of both the object stage and the

K. Steppe et al. / Journal of Structural Biology 148 (2004) 11–21 13

acquisition area on the CCD-camera was used to sup-press ring-artefacts in the reconstructed microCT slices.

Those ring-artefacts are caused by local defects in either

the scintillator or the detector, and they result in faulty

low or high beam intensities (Wildenschild et al., 2002).

Improvement of the signal-to-noise ratio was obtained

by using a 4-frame averaging. This means that the final

X-ray radiograph stored for each of the different angles

resulted from the averaging of four shadow projections(Van Geet et al., 2001). After image acquisition, mi-

croCT serial slices (2-D transverse cross-sections) of the

wood samples were reconstructed with SkyScan�s re-

construction software. This software is based on the

cone-beam reconstruction algorithm of Feldkamp et al.

(1984). During microCT slice reconstruction, a correc-

tion was made to reduce the effects of beam-hardening.

This artefact is caused by an apparent higher X-ray at-tenuation near the periphery of an otherwise homoge-

neous sample. The artefact manifests itself as an

artificial darkening near the edges of the sample (Ket-

cham and Carlson, 2001). Figs. 1A and B show the

images of microCT slices for the beech and the oak

sample, respectively. Finally, the 3-D renderings of the

intact wood samples were built with the software

Fig. 1. (A, B) Oven dry microCT slice of the beech and the oak sample, respe

section of the same beech and oak sample, respectively. The arrow labelled �1�to the cell wall; both are troublesome artefacts associated with microtomy. T

xylem tissue for easy comparison between Figs. 1B and D. Furthermore, GR

package of SkyScan. The X, Y, and Z spatial resolutionwas 10 lm for the three axes.

The physical parameter which is quantified in each

pixel of the microCT slice is the linear attenuation co-

efficient l. The Lambert–Beer law relates the intensity

(I) of X-ray quanta after passing through an object with

thickness h with the incoming intensity (I0). Using the

linear attenuation coefficient (l) for the object, both

intensities are exponentially related:

I ¼ I0 expð�lhÞ: ð1ÞWhen different absorbing materials are present in the X-ray path, a weighted attenuation coefficient (

PlihiÞ can

be used:

I ¼ I0 expð�X

lihiÞ ð2Þ

with index i referring to every type of material occurring

in the X-ray beam.

2.3. Development of image analysis software (lCTanal-ysis)

To perform a reliable analysis of the wood anatomi-

cal characteristics on the microCT slices with computer

ctively. (C, D) Optical micrograph obtained from a stained thin cross-

shows an example of a crack, while the arrow labelled �2� shows damage

he wood structure contained in the white frame highlights a part of the

¼ growth ring; EW ¼ early wood, and LW ¼ late wood.


assistance, the need for self-developed new softwarearose. Although some of the already existing software

packages were able to perform the required measure-

ments, confidence in the results was missing due to the

lack of transparency of the software packages (i.e., used

formulas to calculate certain characteristics were often

not available). As in this study reliable measurements of

the wood anatomical characteristics were of utmost

importance and full control in every step of the imageanalysis was necessary, new software was developed, i.e.,

�lCTanalysis�. This newly developed software automat-

ically detects the objects of interest (i.e., the presence of

xylem vessels on the microCT slices) by a process called

�segmentation�. Various methods are available for seg-

mentation; and for a grey level image, the simplest one is

thresholding. As shown in Figs. 1A–D, the bright areas

represent the lumen of the xylem vessels, whereas darkones represent the remaining part of the wood matrix.

The threshold value defining the boundary between xy-

lem vessels and the wood tissue should then be chosen

somewhere in between both intensity levels. Some false

objects may appear if the threshold value is chosen too

low, or some objects may be missing or may have an

underestimated size if the threshold is chosen too high.

Fig. 2. (C, D) Binary images of the beech and the oak sample created from th

morpho-mathematical filter introduced in the �lCTanalysis� software. The b

vessels (white pixels) and the rest of the wood matrix (black pixels). The imag

micrographs shown in Figs. 1C and D. The data obtained with �lCTanalysi

To overcome these problems, the double thresholdingalgorithm (Soille, 1999) was used; it applies two

threshold values to the image, a high and a low one. The

high threshold selects the objects of interest, but these

selected objects may have incomplete underestimated

areas. The image created using the low threshold value

contains false objects in addition to the true edges of the

objects of interest. By using this algorithm, the incom-

plete areas in the high thresholded image are adjustedand many false objects which were present in the low

thresholded image are eliminated. The resulting binary

microCT slice shows all the vessels of interest, together

with some insuperable artefacts (tiny circular or long

and thin shaped objects). A morpho-mathematical ero-

sion filter with a round structuring element was used to

remove these artefacts (Fig. 2). The maximum thickness

of the artefacts was chosen as diameter of this struc-turing element.

Finally, the xylem vessels were labelled (i.e., they were

given an identification number), after which a quanti-

tative measurement on each individual was processed.

By simply counting the number of pixels inside the vessel

and by multiplying this count with the surface area of

one pixel (the square of the resolution), the surface area

e original images (A, B) using the double threshold algorithm and the

inary images show the distinct boundaries between the lumen of the

es (A) and (B) represent a measurement window selected on the optical

s� is also given.


of each labelled vessel was determined. Assuming anelliptical shape of the vessel lumen, the major and minor

axis were obtained, and were used to calculate the mean

inner diameter (d) of each vessel, defined here as the

mean of the two determined axes. In addition, the po-

rosity was calculated as being the ratio between the total

surface area of all labelled vessel lumina and the total

surface area of the wood matrix viewed on the microCT

image.

2.4. Use of classical microtomy and light microscopy

Optical micrographs of stained thin cross-sections

were recorded from the same wood cubes scanned be-

fore with microCT. Transverse cross-sections of 35 lmthickness were cut on a Microm HM 440 E sliding mi-

crotome using disposable blades (Edge-Rite low profilemicrotome blades), that largely reduce cell wall damage.

The sections were stained with Safranin/Astra blue and

then dehydrated through ethanol solutions of increasing

concentration (30, 50, 70, 96, and 100%). Before

mounting the sections permanently on a slide (with

Entellan), they were stored for one hour in a 1/1 mixture

of clearing agent and alcohol, known as Parasolve

(Jansen et al., 1998). During fixation, one drop ofEntellan per section was used so that the coverslip was

flat, which was necessary for microphotography and

image analysis afterwards.

Optical micrographs of the stained thin cross-sections

(Figs. 1C and D of beech and oak cubes, respectively)

were then recorded with an Olympus AH2 Vanox mi-

croscope system and analysed with a linked digital im-

age analysis system (Soft Imaging System�s analySIS).This commercial software was used to manually count

the total number of vessels within a randomly selected

window on the optical micrograph and to make two

orthogonal measurements of the inner diameter of each

vessel in the selected window. Assuming an elliptical

shape of the vessels, the surface area of each vessel was

determined. The two orthogonal measurements were

used to calculate the mean inner diameter (d) of eachvessel. The ratio between the total surface area of vessel

lumina observed within the window and the total sur-

face area of the selected wood matrix yielded the po-

rosity of the xylem tissue.

2.5. Statistical analysis

Statistical calculations were performed to show that (i)the newly developed �lCTanalysis� software accurately

quantifies anatomical characteristics, and (ii) the mi-

croCT slices can be used for anatomical analysis instead

of optical micrographs of the stained cross-sections. Two

measurement windows were selected randomly on each

optical micrograph: (1518� 1120) lm2 and (1225�903) lm2 for the beech sample, and (1480� 1295) lm2

and (1539� 1196) lm2 for the oak sample. The�lCTanalysis� software (AUTO) then determined vessel

density (total number of vessel/mm2 of transverse cross-

sectional xylem area), the surface area of the individual

vessels, the inner vessel diameter and the porosity. Next,

the commercial software package �analySIS� was used to

manually analyse the same measurement windows

(MAN). The mean (l) and standard deviation (SD) of

each anatomical characteristic for each program (Table 1)were compared with the F-test for equality of two vari-

ances, and the two sample t-test for comparison of means

with either equal or unequal (Satterthwaite approxima-

tion) variances. The sample sizes for both the inner vessel

diameter and the surface area of the vessels were large

enough to satisfy the central-limit theorem (Rosner,

2000). First, the hypothesis H0: lMAN ¼ lAUTO was tes-

ted and accepted for P > 0:05. The second hypothesistested is whether a mean anatomical characteristic ob-

served on either the optical micrograph (lOM) or on a

microCT slice (llCT) are statistically equal (i.e., H0:

lOM ¼ llCT). In each of three and four randomly selected

microCT slices plus the slices shown in Figs. 1A and B,

one measurement window was selected for comparison

with the micrograph of beech and oak. All selected mea-

surement windows (four for beech and five for oak for themicroCT slices, and four for beech and oak for the mi-

crographs) were automatically analysed with �lCTanal-ysis� for the same anatomical characteristics as described

above (Table 2).

3. Results and discussion

3.1. Performance of ‘lCTanalysis’ software for comput-

erized determination of anatomical characteristics

Fig. 2 demonstrates that the use of double thresh-

olds introduced in �lCTanalysis�, together with mor-

pho-mathematical filtering creates accurate binary

images (Figs. 2C and D) of selected measurement

windows as was done, for example, on the opticalmicrographs (Figs. 2A and B). These binary images

show distinct boundaries between the lumen of the

vessel (white pixels) and the rest of the wood matrix

(black pixels).

The inner vessel diameters obtained manually and

with �lCTanalysis� yielded linear determination coeffi-

cients R2 ¼ 0:997 (n ¼ 331) and 0.999 (n ¼ 212) for

beech and oak, respectively; the experimental slopesbeing not significantly different from one. For the sur-

face area of the vessels, R2 ¼ 0:997 (n ¼ 331) and

0.998 (n ¼ 212) for beech and oak, respectively. Like-

wise, none of the mean anatomical characteristics ob-

tained with both measuring techniques (Table 1) differed

significantly. Therefore, �lCTanalysis� offers important

advantages over manual methods in terms of speed and

Table 1

Statistical analysis of inner vessel diameter, vessel surface area, vessel density, and porosity determined from the optical micrographs (see samples shown in Figs. 1C and D) with the manual (MAN)

and lCTanalysis assisted (AUTO) methods

Method Statistic Inner vessel diameter (lm) Vessel surface area (�103lm2) Vessel density (# / mm2) Porosity (%)

Beech Oak Beech Oak Beech Oak Beech Oak

MAN Sample size 331 211 331 211 2 2 2 2

l�SD 32� 10 34� 24 0.87� 0.47 1.31� 1.68 119� 1 56� 15 10.3� 0.3 7.4� 0.6

AUTO Sample size 332 212 332 212 2 2 2 2

l�SD 32� 9 33� 23 0.82� 0.42 1.21� 1.56 119� 4 57� 16 9.7� 0.2 6.8� 0.3

F-test P value 0.072 0.549 0.034 0.299 0.343 0.949 0.586 0.623

H0: r2MAN ¼ r2

AUTO accepted accepted rejecteda accepted accepted accepted accepted accepted

t -test P value 0.396 0.715 0.156a 0.520 0.951 0.981 0.108 0.416

H0: lMAN ¼ lAUTO accepted accepted accepted accepted accepted accepted accepted accepted

The mean (l) and the standard deviation (SD) is given, together with the F-test for equality of variances (r2) and the t-test for the means. The zero hypothesis (H0) is accepted for P values larger

than 0.05.aAs the zero hypothesis for equality of variances was rejected the t-test for samples with unequal variances was performed (Satterthwaite approximation).

Table 2

Statistical analysis of inner vessel diameter, vessel surface area, vessel density, and porosity determined from optical micrographs (OM) and from microCT slices (lCT), using the computer assisted

method (lCTanalysis-software) and the samples shown in Figs. 1C and D and Figs. 1A and B, respectively

Method Statistic Inner vessel diameter (lm) Vessel surface area (�103lm2) Vessel density (# / mm2) Porosity (%)

Beech Oak Beech Oak Beech Oak Beech Oak

OM Sample size 568 541 568 541 4 5 4 5

l�SD 33� 10 35� 25 0.93� 0.47 1.40� 1.76 110� 12 52� 17 10.3� 0.4 7.3� 0.8

lCT Sample size 549 934 549 934 4 4 4 4

l�SD 33� 9 34� 23 0.89� 0.47 1.24� 1.57 108� 15 56� 5 9.7� 1.7 7.2� 0.7

F-test P value 0.739 0.118 1.000 0.002 0.691 0.088 0.051 0.817

H0: r2OM ¼ r2

lCT accepted accepted accepted rejecteda accepted accepted accepted accepted

t-test P value 0.173 0.541 0.135 0.083a 0.883 0.657 0.521 0.766

H0: lOM ¼ llCT accepted accepted accepted accepted accepted accepted accepted accepted

The mean (l) and the standard deviation (SD) is given, together with the F-test for equality of variances (r2) and the t-test for the means. The zero hypothesis (H0) is accepted for P values larger

than 0.05.aAs the zero hypothesis for equality of variances was rejected the t-test for samples with unequal variances was performed (Satterthwaite approximation).

16

K.Step

peet

al./JournalofStru

cturalBiology148(2004)11–21


simplicity. Moreover, measurement errors made by hu-man interpretation errors (through operator fatigue) are

also excluded.

Particular attention has to be paid to the standard

deviations (SD) calculated for the inner vessel diameter

(d) and the surface area of the vessels (A) (Table 1). Thecoefficient of variation (CV) for d, defined as the ratio

between SD and the mean value (l), yields a value of 0.3

and 0.7, and the CV for A of 0.5 and 1.28 for beech andoak, respectively. The smaller CV of beech for both dand A compared to oak can be explained in terms of

wood anatomical differences between both tree species.

The beech sample is a typical example of diffuse-porous

wood. Therefore, the xylem vessels are fairly uniform in

distribution and size throughout the entire growth ring.

The oak sample, however, is a typical example of ring-

porous wood characterised by size differences of thevessels in a growth ring, whereby the vessels of the early

wood are larger than those found in the late wood

(Schweingruber, 1990). This is confirmed by the histo-

grams of the inner vessel diameter distribution measured

within the randomly selected windows on the optical

micrographs (Section 2.5) as shown in Figs. 3A and B.

The histogram of beech was indeed more symmetric,

whereas the one for oak was positively skewed. More-over, Fig. 3A illustrates that in the beech histogram, a

Fig. 3. Histogram of the inner vessel diameter distribution of the beech

sample (A) and the oak sample (B). The inner vessel diameters (d) weremeasured within the randomly selected windows on the optical mi-

crograph of Figs. 1C and D, respectively (see also text).

major fraction of the inner vessel diameters lay veryclose to the calculated mean of 32 lm (Table 1). On the

contrary, the histogram presented in Fig. 3B shows that

the inner vessel diameters of the oak sample in the

skewed distribution ranged from small late wood vessels

to large early wood vessels. Although the mean vessel

diameter for oak (l ffi 34 lm—Table 1) was almost

equal to the one for beech, its higher CV was due to the

more extended range of the inner vessel diameterscompared to the beech sample.

Noteworthy is also the fact that the transverse cross-

sections depicted in Fig. 1 reveal a xylem anatomy that

is slightly different from the one that is expected for

mature trees. Whereas the seasonal trend in xylem

structure for the young beech and oak tree corresponds

well with that normally observed for large trees (i.e., a

large early vessel size and a marked seasonal change invessel diameter for the ring-porous oak tree, in contrast

with a smaller maximum vessel size and little seasonal

change in vessel diameter for the diffuse-porous beech

tree), the vessel sizes observed were smaller than those

found in mature xylem. For mature wood of oak, vessel

diameters are reported to range from 20 to 30 lm for

late wood vessels up to 300–500 lm for early wood

vessels. For mature wood of beech trees, vessel diame-ters generally range between 50 to 85 lm (Grosser, 1977;

Jacquiot et al., 1973; Wagenf€uhr and Scheiber, 1989).

However, for the young oak wood the vessel diameters

were found to vary between the limits of 10 and 90 lm,

while for the young beech wood diameters ranged be-

tween 10 and 55 lm (Fig. 3). Considering the fact that

xylem elements typically become longer and wider with

increased cambial age or stem diameter (Gartner, 1995;Mencuccini et al., 1997), these results could be expected.

3.2. Performance of X-ray microCT versus optical

micrographs

Typical microCT slices for both the beech and the

oak sample are shown in Figs. 1A and B respectively.

The bright pixels represent the lumen of the xylem ves-sels, whereas the dark ones represent the rest of the

wood matrix (i.e., fibres, parenchyma cells, and wood

rays). As each pixel of the microCT slice quantifies the

linear attenuation coefficient (l), density differences of

the xylem tissue are revealed, whereby dark pixels result

from high X-ray attenuation and correspond with high

wood density. The dark top part of Fig. 1A reveals the

high density of the late wood in contrast with the lightearly wood of the subsequent growth ring. In Fig. 1B, a

growth ring can be distinguished with clear density dif-

ferences between the early and the late wood. Computed

tomography has been shown to be capable of deter-

mining with high precision the wood density in a non-

destructive way (e.g., Fromm et al., 2001; Lindgren,

1991). However, calculation of such wood densities

Fig. 4. (A, B) Wet microCT slice of the same beech and oak sample as

shown in Figs. 1A and B, respectively. The beech and the oak sample

were scanned in wet condition immediately after removal from the

glycerine–ethanol solution. Only the non-filled vessels are visible. The

arrows labelled �1� indicate the lack of (small) vessels in the late wood; the

arrows labelled �2� show the so-called noise which is caused by the �in-visible� vessels trying to manifest themselves (see text). The white square

highlights a part of the xylem tissue for easy comparison with Fig. 1B.


using the microCT slices was beyond the scope of thisstudy.

Visual inspection of the microCT images reveals a

very good correspondence with the micrographs ob-

tained from stained thin sections (Figs. 1C and D). This

is highlighted in Fig. 1 by the wood structure contained

in the white frame. Despite the fact that the pixel size of

1.7� 1.7 lm2 in the optical micrographs is smaller than

that in the microCT slices (pixel size ¼ 7.0� 7.0 lm2),visual perception suggested that the latter can be

substituted for optical micrographs. All vessels larger

than 10 lm in size could be successfully detected by

microCT. Some authors (Kinney and Nichols, 1992;

Ritman et al., 1997; Stock et al., 2003a; Thurner et al.,

2003) already demonstrated that images obtained from

microCT indeed were closely correlated to optical mi-

crographs of the objects under study. Lane et al. (2003),as well as others, showed that it is possible to derive

similar histological quantities (i.e., relative bone volume,

trabecular number, and trabecular thickness) from ei-

ther microCT images or thin histological cross-sections

cut with a microtome and analysed with light micros-

copy. The statistical proof of the similarity for the case

of wood samples in this study (Table 2) is very impor-

tant as all the major drawbacks and limitations linkedwith microtomy can eventually be overcome by using

microCT for wood anatomical research (Bentley et al.,

2002). The complex and time-consuming procedure for

obtaining good wood slices from microtomy (Jansen

et al., 1998) (going from softening of the wood sample,

over cutting and treating the cross-sections to mounting

them) can now be avoided as microCT allows non-in-

vasive mapping of 2-D slices without sample prepara-tion or chemical fixation (SkyScan, 1998). As such, no

cracks (see arrow labelled �1� in Fig. 1D) and no cell wall

damage (see arrow labelled �2� in Fig. 1D), both trou-

blesome artefacts associated with microtomy, have to be

expected any longer. Moreover, the spatial resolution of

microCT slices, already approaching that of optical

microscopy (Stock et al., 2003b), can be further im-

proved up to 1 lm or better by using synchrotron ra-diation microCT (Thurner et al., 2003). The main

difference with the system used here is only the type of

X-ray source. In contrast with the polychromatic beam

produced by conventional X-ray sources, the synchro-

tron type has the capability to select tuned monochro-

matic (i.e., single energy) X-rays (M€uller et al., 2002;

Wildenschild et al., 2002). Objects of interest with

dimensions less then 10 lm should be imaged withsynchrotron radiation microCT instead of conven-

tional X-ray sources. For the purpose of this study, the

spatial resolution of the SkyScan was sufficient as the

typical inner vessel diameter of the wood samples was

larger than 10 lm.

However, to obtain such a good correspondence be-

tweenmicroCT slices and optical micrographs oven dried

samples had to be used for the microCT method. It wasfound that, after being thoroughly rinsed with ethanol to

remove the glycerine–ethanol mixture from the vessels,

drying of the samples could be performed at a tempera-

ture of 50 �C without modifying the dimensions of the

wood cubes. Figs. 4A and B illustrate the microCT slices

Fig. 5. (A, B) MicroCT derived 3-D rendering of the beech and the oak

sample, respectively, showing the complex internal vessel network.

Both indicated scale bars equal 250lm.


of the same beech and oak sample, but scanned under�wet� conditions immediately after removal from the

glycerine–ethanol mixture. This procedure revealed that

less visible vessels could be detected (only half the amount

of vessels per unit xylem area counted on the optical mi-

crograph of beech and only one fourth the amount

counted on the optical micrograph of oak). Due to strong

capillary forces associated with small dimensions, the

vessels having the smallest diameter remained completelyfilled (see arrows labelled �1�). Noteworthy is that the

apparent noise observed in the images of Figs. 4A and B

(see arrows labelled �2�) was not due to errorsmade during

the reconstruction of the images. This so-called �noise�should be attributed to �invisible� vessels which are only

partly expressed by X-ray attenuation. The lack of visi-

bility and occurrence of blurring in the samples scanned

under wet conditions can be explained by the fact thatthere was almost no difference in X-ray absorption be-

tween the with glycerine–ethanol solution filled vessels

and the sample itself.

3.3. Application of microCT for 3-D renderings of xylem

anatomy

The microCT also provides a possibility to reveal anobject�s internal microstructure in a 3-D space. It has

been demonstrated as an interesting tool for 3-D imag-

ing of various materials, and for a wide range of bio-

logical applications such as sea urchin teeth (Stock et al.,

2003a,b), bone tissue (Borah et al., 2001; Lane et al.,

2003) and various basic functional units within intact

animal organs, such as the renal glomeruli and their

postglomerular vessels, the Haversian canals in bone,cardiac coronary branching structure, and myocardial

muscle fibre architecture (Bentley et al., 2002; Ritman

et al., 1997). In this study, the internal 3-D microstruc-

ture was reconstructed for both the beech and oak

sample. The 3-D renderings of the intact wood samples

shown in Figs. 5A and B were produced by stacking the

sequential microCT slices with SkyScan�s proprietary

software package. The two images illustrate the abilityto spatially isolate individual vessels within the complex

wood matrix. It can be seen that the vessels do not run

in a neat parallel way, but that they deviate from their

axial path to form a complex network. Some of the

vessels are jointed together by intervessel bordered pits,

while others drift apart (see Fig. 5A). The physiological

significance of this networking in the wood matrix is the

tangential spreading of the axial path of water transportin living trees (Tyree and Zimmermann, 2002). This

implies that water taken up by any root ascends in the

trunk while spreading out laterally through the growth

rings. Water supplied by a particular root not only

reaches a particular branch but is redistributed over a

large part of the crown. In other words, each branch of

the crown can obtain its water from many different

roots. This can be considered as an important safety

feature for living trees: loss of one or more roots does

not impede the water supply of individual branches.

Such a 3-D network of vessels was described first by

Braun (1959). He meticulously reconstructed the vesselnetwork of Populus from a series of transverse micro-

scopic sections. As his method was time-consuming and

impractical, Zimmermann and Tomlinson (1966) intro-

duced a cinematographic method. This simplified the

visualisation of the 3-D vessel structure to some extent

by assembling, frame-by-frame, a 16-mm motion picture

film from series of cross-sections photographed in the

microscope. However, this cinematographic method stillremains very labour-intensive and, as such, it has been

hardly used in recent years. Another and easier method,

i.e., the resin-casting technique (Fujii, 1993; Kitin et al.,


2001), was developed which visualizes the vessel luminaof a wood sample in the form of silicon filaments. These

silicone casts are made by perfusing the vessels with

polystyrene and after polymerisation of the polystyrene,

by removing all the organic material with acidic solu-

tions and/or enzymes. Although it provides high spatial

resolution, this method also suffers from some problems

which are related to the dislocation of resin casts during

the dissolution of cell wall materials, and the deforma-tion of resin casts in samples with low packing density

(Fujii, 1993). At this point, it becomes clear that the

microCT approach overcomes many of the problems

mentioned above. Illman and Dowd (1999) successfully

applied for the first time microCT to characterise the 3-

D geometry of the wood matrix of sound pine wood and

demonstrated the feasibility of microCT to quantify the

microscale breakdown of the tracheids by brown-rotwood decay fungi. Furthermore, as microCT is a non-

invasive technique, the sample remains available for

further investigations. As illustrated in Figs. 5A and B,

the present study also revealed that the spatial resolu-

tion of the 3-D renderings was insufficient to resolve the

fine details of the intervessel microstructures (i.e., bor-

dered pits) as is the case for the resin-casting technique.

Again, the synchrotron radiation microCT (see above)should be used to increase the spatial resolution for vi-

sualization of such microstructure details.

4. Conclusions

The time-consuming and difficult procedure to obtain

cross-sections of wood samples by microtomy (e.g.,problems of cross-section curling and damage due to

cutting), makes use of optical micrographs often im-

practical to those who need quantitative data of wood

anatomical characteristics in a short period of time.

Therefore, a novel approach based on X-ray computed

microtomography (microCT) was tested against the

drawbacks and limitations of classical microtomy. As a

first step, new software (lCTanalysis) was developed toimprove image processing of 2-D images in terms of

speed and simplicity. The statistical analysis on the re-

sults obtained with �lCTanalysis� software revealed that

the new software permits accurate and fast measurement

of vessel characteristics of wood samples.

Visual inspection slices produced with microCT re-

vealed a very good correspondence with the micro-

graphs of the stained thin sections for both beech andoak samples. This close correspondence was also con-

firmed by statistical analysis. Therefore, it was argued

that use of micrographs can be substituted by microCT

slices without any loss of accuracy. It was found, how-

ever, that the wood samples used for microCT should be

oven dry. Wet samples, which still contain the glycerine–

ethanol mixture, yield much less contrast and image

detail is lost. Moreover, microCT also provides thepossibility to explore the internal 3-D microstructure of

the wood matrix. By stacking sequences of microCT

slices with proprietary software of the instrument, 3-D

renderings can be produced, showing the complexity of

the vessel network in both the beech and the oak sample.

In conclusion, this study has demonstrated that mi-

croCT in combination with adequate software can suc-

cessfully be used when quantitative data of woodanatomical characteristics are needed in either 2-D or in

3-D space. The method provides information on vessel

diameter, vessel area, vessel density, and porosity with

the same accuracy as classical microtomy.

Acknowledgments

The authors wish to thank the Fund for Scientific

Research–Flanders (Belgium) (F.W.O.-Vlaanderen) and

the Institute for the Promotion of Innovation by Science

and Technology in Flanders (Belgium) for the Ph.D.

fundings granted to the first and the second author,

respectively. We also are grateful to the Laboratory of

Wood Technology (Ghent University–Belgium; head:Prof. Dr. ir. M. Stevens) for the offered possibility to

practise classical microtomy.

References

Atkinson, C.J., Taylor, J.M., 1996. Effects of elevated CO2 on stem

growth, vessel area and hydraulic conductivity of oak and cherry

seedlings. New Phytol. 133, 617–626.

Bentley, M.D., Ortiz, M.C., Ritman, E.L., Romero, J.C., 2002. The

use of microcomputed tomography to study microvasculature in

small rodents. Am. J. Physiol. Reg. I. 282, R1267–R1279.

Bentz, D.P., Quenard, D.A., Kunzel, H.M., Baruchel, J., Peyrin, F.,

Martys, N.S., Garboczi, E.J., 2000. Microstructure and transport

properties of porous building materials. II: three-dimensional X-

ray tomographic studies. Mater. Struct. 33, 147–153.

Borah, B., Gross, G.J., Dufresne, T.E., Smith, T.S., Cockman, M.D.,

Chmielewski, P.A., Lundy, M.W., Hartke, J.R., Sod, E.W., 2001.

Three-dimensional microimaging (MRlI and lCT), finite element

modeling, and rapid prototyping provide unique insights into bone

architecture in osteoporosis. Anat. Rec. 265, 101–110.

Braun, H.J., 1959. Die vernetzung der gef€asse bei populus. Z. Bot. 47,

421–434.

Cnudde, V., Jacobs, P., 2003. Preliminary results of X-ray micro-

tomography applied in conservation and restoration of natural

building stones, in: Proceedings of the International Workshop on

X-ray CT for Geomaterials. Soils, Concrete, Rocks, Kumamoto,

Japan, pp. 329–336.

Falcon-Lang, H.J., 2000. A method to distinguish between woods

produced by evergreen and deciduous coniferopsids on the basis of

growth ring anatomy: a new palaeoecological tool. Palaeontology

43, 785–793.

Feldkamp, L.A., Davis, L.C., Kress, J.W., 1984. Practical cone-beam

algorithm. J. Opt. Soc. Am. A. 1, 612–619.

Fromm, J.H., Sautter, I., Matthies, D., Kremer, J., Schumacher, P.,

Ganter, C., 2001. Xylem water content and wood density in spruce


and oak trees detected by high-resolution computed tomography.

Plant Physiol. 127, 416–425.

Fujii, T., 1993. Application of a resin casting method to wood

anatomy of some Japanese Fagaceae species. IAWA J. 14, 273–

288.

Gartner, B.L., 1995. Patterns of xylem variation within a tree and their

hydraulic and mechanical consequences. In: Gartner, B.L. (Ed.),

Plant Stems: Physiology and Functional Morphology. Academic

Press, San Diego, pp. 125–149.

Gartner, B.L., Roy, J., Huc, R., 2003. Effects of tension wood on

specific conductivity and vulnerability to embolism of Quercus ilex

seedlings grown at two atmospheric CO2 concentrations. Tree

Physiol. 23, 387–395.

Grosser, D., 1977. Die H€olzer Mitteleuropas. Ein microphotographi-

scher Lehratlas. Springer-Verlag, Berlin.

Hoffmann, S., Schweingruber, F.H., 2002. Light shortage as a

modifying factor for growth dynamics and wood anatomy in

young deciduous trees. IAWA J. 23, 121–141.

Illman, B.L., Dowd, B.A., 1999. High resolution microtomography for

density and spatial information about wood structures. In: Bonse,

U. (Ed.), Proceedings of SPIE on Developments in X-ray Tomog-

raphy II, vol. 3772. Society of Photo-Optical Instrumentation

Engineers, Washington, pp. 198–204.

Jacquiot, C., Trenard, Y., Diral, D., 1973. Atlas d�anatomie des bois

des angiospermes. Centre Technique du Bois, Paris.

James, S.A., Meinzer, F.C., Goldstein, G., Woodruff, D., Jones, T.,

Restom, T., Mejia, M., Clearwater, M., Campanello, P., 2003.

Axial and radial water transport and internal water storage in

tropical forest canopy trees. Oecologia 134, 37–45.

Jansen, S., Kitin, P., de Pauw, H., Idris, M., Beeckman, H., Smets, E.,

1998. Preparation of wood specimens for transmitted light

microscopy and scanning electron microscopy. Belg. J. Bot. 131,

41–49.

Ketcham, R.A., Carlson, W.D., 2001. Acquisition, optimization and

interpretation of X-ray computed tomographic imagery: applica-

tions to the geosciences. Comput. Geosci. 27, 381–400.

Kinney, J.H., Nichols, M.C., 1992. X-ray tomographic microscopy

(XTM) using synchrotron radiation. Annu. Rev. Mater. Sci. 22,

121–152.

Kitin, P., Sano, Y., Funada, R., 2001. Analysis of cambium and

differentiating vessel elements in Kalopanax pictus using resin cast

replicas. IAWA J. 22, 15–28.

Lane, N.E., Yao, W., Kinney, J.H., Modin, G., Balooch, M., Wronski,

T.J., 2003. Both hPTH(1-34) and bFGF increase trabecular bone

mass in osteopenic rats but they have different effects on trabecular

bone architecture. J. Bone Miner. Res. 18, 2105–2115.

Lindgren, L.O., 1991. Medical CAT-scanning: X-ray absorption

coefficients, CT-numbers and their relation to wood density. Wood

Sci. Technol. 25, 341–349.

Lovisolo, C., Schubert, A., 1998. Effects of water stress on vessel size

and xylem hydraulic conductivity in Vitis vinifera L. J. Exp. Bot.

49, 693–700.

Mencuccini, M., Grace, J., Fioravanti, M., 1997. Biomechanical and

hydraulic determinants of tree structure in Scots pine: anatomical

characteristics. Tree Physiol. 17, 105–113.

M€uller, B., Beckmann, F., Huser, M., Maspero, F., Sz�ekely, G.,

Ruffieux, K., Thurner, P., Wintermantel, E., 2002. Non-destructive

three-dimensional evaluation of a polymer sponge by micro-

tomography using synchrotron radiation. Biomol. Eng. 19, 73–78.

Perret, J., Prasher, S.O., Kantzas, A., Langford, C., 1999. Three-

dimensional quantification of macropore networks in undisturbed

soil cores. Soil Sci. Soc. Am. J. 63, 1530–1543.

Ritman, E.L., Jorgensen, S.M., Lund, P.E., Thomas, P.J., Dunsmuir,

J.H., Romero, J.C., Turner, R.T., Bolander, M.E., 1997. Synchro-

tron-based micro-CT of in situ biological Basic Functional Units

and their integration. In: Bonse, U. (Ed.), Proceedings of SPIE on

Developments in X-ray Tomography, vol. 3149. Society of Photo-

Optical Instrumentation Engineers, Washington, pp. 13–24.

Rosner, B., 2000. Fundamentals of Biostatistics, fifth ed. Duxbury,

USA.

Schultz, H.R., Matthews, M.A., 1993. Xylem development and

hydraulic conductance in sun and shade shoots of grapevine (Vitis

vinifera L.)–evidence that low light uncouples water transport

capacity from leaf area. Planta 190, 393–406.

Schweingruber, F.H., 1990. Anatomy of European woods. An atlas for

the identification of European trees, shrubs and dwarf shrubs. Paul

Haupt Berne and Stuttgart Publishers, Bern.

SkyScan, 1998. SkyScan 1072 desktop X-ray microtomograph:

instruction manual. SkyScan, Aartselaar.

Soille, P., 1999. Morphological Image Analysis. Principles and

Applications. Springer-Verlag, Berlin.

Steppe, K., Lemeur, R., 2004. An experimental system for analysis of

the dynamic sap flow characteristics in young trees: results of a

beech tree. Funct. Plant. Biol. 31, 83–92.

Stock, S.R., Ignatiev, K.I., Dahl, T., Veis, A., De Carlo, F., 2003a.

Three-dimensional microarchitecture of the plates (primary, sec-

ondary, and carinar process) in the developing tooth of Lytechinus

variegatus revealed by synchrotron X-ray absorption microtomog-

raphy (microCT). J. Struct. Biol. 144, 282–300.

Stock, S.R., Nagaraja, S., Barss, J., Dahl, T., Veis, A., 2003b. X-ray

microCT study of pyramids of the sea urchin Lytechinus variegatus.

J. Struct. Biol. 141, 9–21.

Thurner, P., M€uller, B., Beckmann, F., Weitkamp, T., Rau, C.,

M€uller, R., Hubbell, J.A., Sennhauser, U., 2003. Tomography

studies of human foreskin fibroblasts on polymer yarns. Nucl.

Instrum. Meth. B 200, 397–405.

Tyree, M.T., Zimmermann, M.H., 2002. Xylem Structure and the

Ascent of Sap, second ed. Springer-Verlag, Berlin.

Van Geet, M., Swennen, R., Wevers, M., 2001. Towards 3-D

petrography: application of microfocus computer tomography in

geological science. Comput. Geosci. 27, 1091–1099.

Wagenf€uhr, R., Scheiber, C., 1989. Holzatlas 3. Auflage. VEB

Fachbuchverlag, Leipzig.

Wildenschild, D., Hopmans, J.W., Vaz, C.M.P., Rivers, M.L., Rikard,

D., Christensen, B.S.B., 2002. Using X-ray computed tomography

in hydrology: systems, resolutions, and limitations. J. Hydrol. 267,

285–297.

Zimmermann, M.H., Tomlinson, P.B., 1966. Analysis of complex

vascular systems in plants: optical shuttle method. Science 152, 72–

73.

Use of X-ray computed microtomography for non-invasive determination of wood anatomical...

Documents

Transcript of Use of X-ray computed microtomography for non-invasive determination of wood anatomical...