Biotechnol. Appl. Biochem.
(2004) 39, 49–58 (Printed in Great Britain)
Computer-aided tissue engineering: application to biomimetic
modelling and design of tissue scaffolds
Wei Sun1 , Binil Starly, Andrew Darling and Connie Gomez
Laboratory for Computer-Aided Tissue Engineering, Department of Mechanical Engineering and Mechanics,
College of Engineering, Drexel University, 3141 Chestnut Street, Philadelphia, PA 19104, U.S.A.
Computer-aided tissue engineering (CATE) enables
many novel approaches in modelling, design and fabri-
cation of complex tissue substitutes with enhanced
functionality and improved cell–matrix interactions.
Central to CATE is its bio-tissue informatics model
that represents tissue biological, biomechanical and
biochemical information that serves as a central re-
pository to interface design, simulation and tissue
fabrication. The present paper discusses the application
of a CATE approach to the biomimetic design of bone
tissue scaffold. A general CATE-based process for
biomimetic modelling, anatomic reconstruction, com-
puter-assisted-design of tissue scaffold, quantitative-
computed-tomography characterization, finite ele-
ment analysis and freeform extruding deposition for
fabrication of scaffold is presented.
Introduction
Computer-aided tissue engineering (CATE) is a newly
emerging field that can be classified within three major
categories: (1) computer-aided tissue modelling; (2) com-
puter-aided tissue informatics; and (3) computer-aided tissue
scaffold design and manufacturing ([1]; our review in this
issue [2]). Application of CATE allows us to explore
many novel ideas in modelling, design and fabrication of
tissue scaffolds with enhanced functionality and improved
interactions with cells. This is particularly useful in modelling
and design of complex bone tissue scaffolds and replacement
structures that require us to simultaneously consider
many biological and biophysical design requirements [3–5].
Generation of functional tissue or organ structure requires
a scaffold to guide the overall shape and three-dimensional
organization of multiple cell types. The internal architecture
of a bone scaffold cannot be chosen at ease, and various
factors need to be considered, such as porosity, pore size
and interconnectivity of the tissue scaffold structure, have
all been identified to be important factors that make a tissue
scaffold successful. These factors aid in the transportation
of nutrients that would enable the growth of new cells
and allow the tissue scaffold to act as a suitable template
49
for appropriate bone ingrowth and healing. In addition, the
designed scaffold structure should be able to have required
mechanical strength after implantation, particularly in the
reconstruction of hard and load-bearing tissues such as bone
and cartilage. The strength of biodegradable tissue scaffolds
must not decline rapidly and must degrade at a rate similar to
the growth of new tissue cells. Therefore the tissue scaffold
structure should be compliant, both biologically as well as
mechanically, at the site of implantation and should at best
mimic the natural bone properties in order to function as a
true bone substitute.
Central to CATE approach is its ability to represent
pertinent tissue biological, biomechanical and biochemical
information as a computer-based bio-tissue informatics
model. This model can be used as an effective communi-
cation tool between biologists and tissue engineers, with
the database of the model serving as a central repository
to interface design, simulation and manufacturing of tissue
substitutes. An overview of CATE in terms of the three
major categories, particularly in description of the meth-
odology of using high-resolution non-invasive imaging for
CATE model generation, image-based three-dimensional
(3D) reconstruction, computer-aided tissue informatics,
tissue scaffold design, freeform fabrication and bio-blueprint
modelling and its application to 3D organ printing are
presented in our review in this issue [2]. The objective of
the present paper is to discuss the application of the CATE
approach for a biomimetic design of bone tissue scaffold.
The outline of the developed biomimetic approach with
its application in modelling heterogeneous bone structure,
quantitative computed tomography characterization of bone
tissue properties, and various modeling alternatives to
Key words: bone scaffold, computer-aided design (CAD), computer-aided
tissue engineering (CATE), tissue engineering, tissue scaffold design.
Abbreviations used: CATE, computer-aided tissue engineering; CAD,
computer-aided design; CT, computed tomography; FEA, fine element
analysis; IGES, initial graphics exchange specification; MRI,
magnetic-resonance imaging; PLGA, poly(DL-lactic-co-glycollic acid); PLLA,
poly(L-lactic acid); QCT, quantitative computed tomography; NURBS,
non-uniform rational B-spline; STEP, standard for exchange of product
model data; STL, stereolithography; 3D, three-dimensional.
1
To whom correspondence should be addressed (e-mail
sunwei@drexel.edu).


C 2004 Portland Press Ltd
50 W. Sun and others
construct a CAD (computer-aided design)-based tissue
heterogeneous model will be introduced in the section
entitled ‘CATE based biomimetic modelling of bone
scaffolds’. The design and analysis of different tissue
representative unit cells and its utilization as building blocks
to assemble the designed tissue scaffold will be presented
in the section entitled ‘Design and characterization of bone
scaffold and tissue representative unit cells’, followed by
an introduction to using an in-house developed fabrication
planning and extruding deposition method to freeform fabri-
cate the tissue scaffold in the section entitled ‘Freeform
fabrication of tissue scaffold’ and a summary, discussion and
conclusions in the final section (‘Conclusions’).
CATE-based biomimetic modelling
of bone scaffolds
Bone tissue scaffolds need to have certain characteristics of
their own in order to function as a true bone substitute
that satisfies the biological, mechanical and geometrical
constraints. Such characteristics include: (1) biological
requirements – the designed scaffold must facilitate cell
attachment and distribution, growth of regenerative tissue
and facilitate the transport of nutrients and signals. This
requirement can be achieved by controlling the porosity
of the structure, by providing appropriate interconnectivity
inside the structure, and by selecting suitable biocompatible
materials; (2) mechanical requirement – the designed scaf-
fold must provide structural support at the site of replace-
ment while the tissue regenerates to occupy the space
defined by the scaffold structure. Scaffold structures need
to be defined to possess the required mechanical stiffness
and strength of the replaced structure; and (3) anatomical
requirement – it must be of an appropriate geometric size
that fits in at the site of replacement [6–8].
Our overall CATE design approach begins with the
acquisition of non-invasive images and image processing
of appropriate tissue region of interest. This is followed
by a 3D reconstruction of anatomical structure using
commercially available medical reconstructive and reverse
engineering software (MIMICS [9] and Geomagic [10]).
The next step is to appropriately devise a process by
which tissue structural heterogeneity can be characterized
through a homogenization technique, and to define tissue
anatomic representative features for generating CAD
(Pro/ENGINEER
C
[11])-based unit cells. On the basis of the
designed CAD geometrical configuration and the intended
scaffolding materials, a finite element method (ABAQUS
[12]) is applied to determine the corresponding mechanical
properties. Those properties are further compared with
the replaced-tissue mechanical properties characterized by


C 2004 Portland Press Ltd
the QCT (qualitative computed tomography) method. The
unit cells with matching properties are selected as candi-
date unit cells to make up the tissue scaffold. The candidate
unit cells will be further evaluated according to their internal
architectures and the intended biological purpose. Using
CAD solid-modelling-based Boolean operations, a set of
such selected unit cells would be integrated with the shape
of the bone to form the bone tissue scaffold with specified
internal architecture and structural properties to match that
of the actual replaced bone based on the characterization
analysis. Once the complete CAD database of the bone
tissue scaffold structure is in place, process planning and
toolpath instructions will be generated based on solid-
freeform-fabrications techniques, which would be able to
manufacture the designed tissue structures. The overall
procedure for the CATE-based biomimetic modelling and
design of the bone tissue scaffold is illustrated in Figure 1
and further described in the following steps.
Step 1: image acquisition
Our region of study involved the CT (computed tomo-
graphy) images obtained from that of a proximal femur bone.
In all, 34 sliced images were obtained each, of 2 mm slices,
thereby totalling a length of 68 mm of the proximal femur
bone. These sliced images were loaded into MIMICS and
organized in a sequential manner and oriented for its top,
bottom, anterior and posterior positions.
Step 2: segmentation and characterization process
Once loaded, the region of interest was identified and a 3D
voxel model of the bone under study had to be made. As
a first step, an appropriate threshold range was found that
could best capture the relevant information contained in
the femur. Using this threshold value, all pixels within this
range were grown to a colour mask and hence the segmen-
tation process achieved by making use of region-growing
techniques available in the software. The segmentation was
achieved using two different approaches, each approach
serving a particular purpose. In the first approach, the
whole proximal femur structure was grown into one
single colour mask representing one single threshold range.
With this approach, the average threshold value for the
whole structure could be obtained. This average threshold
value was in turn correlated to the quantitative computed
tomography number (QCT#) represented in Hounsfield
units (HU) by using a simple relation as follows:
QCT# = threshold value − 1024 (1)
In the second approach, described as the homogenization
technique, the femur structure was divided into layers and
then an average QCT# for each layer found. A collection
of slices of the femur was grouped as layers and segmented
 Computer-aided tissue engineering: application to biomimetic modelling and design of tissue scaffolds
Figure 1 Overall procedures of modelling and design of biomimetic bone scaffold
Figure 2 Average QCT values measured from CT
using different colour masks. Each layer thickness was
about 10 mm and there were about seven layers in all.
An average QCT# was obtained for each layer in order to
characterize the tissue heterogeneity (Figure 2). The QCT#
retrieved from the appropriate layers is then correlated
with the density (ρ) of the bone by a linear interpolation
51
using relations available in published literature. This density
can in turn be then related to elasticity (E), allowing the
heterogeneous elasticity of the bone to be defined through
the relations obtained as described in [13,14]:
For QCT < 816 : ρ = 1.9 × 10−3 · QCT + 0.105 and
E = 0.06 + 0.9 · ρ 2 (2)
For QCT < 816 : ρ = 7.69 × 10 + 1.028 and −4
E = 0.09 · ρ 7.4 (3)
The structural heterogeneity of the bone can thus be
defined through the associated bone Young’s modulus. The
characterization results are as shown in Table 1. The last
row in the Table indicates the QCT# retrieved when a
single colour mask was considered. The calculated E is in
accordance with published data for cancellous bone, namely
0.5–1.5 GPa. It is noteworthy that both cancellous and
cortical bone have been considered smeared together as
one structure in each layer and hence the slightly higher
values obtained by this technique.


C 2004 Portland Press Ltd
52 W. Sun and others

Table 1 Bone spatial heterogeneity: seven-layer model tissue model using three different interfaces with CT/MRI
Layer no. QCT# 2
Density (g/cm ) E (MPa)
(magnetic-resonance imaging) data: (1) MedCAD interface
(direct interface with medical-imaging data within MIMICS
1 728 1.4882 2053 [9]); (2) reverse-engineering interface (reconstruction from
2 603 1.2507 1467
3 479 1.0151 987 point to curve to surface to solid using Geomagic [11]);
4 385 0.8365 690 and (3) STL (stereolithography) interface (using STL from
5 374 0.8156 658 MIMICS). The process flow chart of the three approaches is
6 390 0.8460 704
7 359 0.7871 617 presented in Figure 3.
Entire bone 474 1.0056 970
Process path 1: MedCAD interface This interface bridges the
gap between medical imaging and CAD design software,
meaning that this interface can export data from the imaging
Step 3: CAD model generation system to the CAD system and vice versa. The IGES (initial
Once the bone structure has been characterized, a complete graphics exchange specification) CAD file transfer format
CAD database of the structure needs to be generated. We is used for models to be exported to CAD and the STL
conducted the 3D reconstruction to arrive at a CAD-based format used to import models from CAD to the MedCAD

Figure 3 Flowchart for imaging-based CAD model generation

C 2004 Portland Press Ltd
 Computer-aided tissue engineering: application to biomimetic modelling and design of tissue scaffolds
Figure 4 Reverse-engineering approach for developing a CAD-based femur model
system. It can be seen that not all features or details of the
image can be exported to CAD, which would otherwise
slow the system down. Basic features can be recognized and
converted into geometrical entities such as cylinder, sphere,
anatomical landmarks etc. B-Spline surfaces are also used to
make a real fit on a surface anatomical contour.
Process path 2: reverse-engineering method The 3D voxel
model created after segmentation is used as the starting
point for this method. The 3D voxel dataset of the bone
structure is converted into point data form and then these
points are loaded into any reverse-engineering software. The
commercially available software Geomagic Studios [11] was
then used to process these data points for surface processing
and refinement. Application of this process path to develop
a CAD-based femur model is outlined in Figure 4. This is
perhaps the best approach that can be followed, since the
process starts from the base level, i.e. points. Although
the model has a lot of processing time involved, the results
obtained are significantly better than those obtained by the
other two methods.
Process path 3: the STL interface The 3D voxel model
can also be converted into the STL file and this STL file can
then be imported into Geomagics for surface refinement
and non-uniform rational B-spline (NURBS) surface gene-
ration. The difference between this and the point data
method is that here a triangulated model is the starting
input format rather than the point data form of the model
structure. The process time involved is less but may be
suitable for only certain kinds of surfaces.
53
The final CAD model of the proximal femur structure
obtained by each of the process path is shown in Figure 5.
Notice that the femur structure obtained by the STL process
path has resulted in a poor CAD model. The smoother the
model, the lower the file sizes tend to be, and this can
be less frustrating when handling these models in CAD
software. The reverse-engineering method was followed
for the CAD model generation of the homogenized femur
domain structure, since this resulted in a much smoother
and better model and, more importantly, had a smaller
CAD file size of around 10 MB for all the seven layers
combined. The seven-layered CAD assembly structure of
the proximal femur is also shown in Figure 5. A summary
of the comparison of each process path is presented in
Table 2.
Design and characterization of bone
scaffold and tissue representative
unit cells
Cellular scaffold unit cells with various porosities and inter-
nal architectures can provide tailored mechanical and biolo-
gical properties. Here we use feature primitives in which
each primitive discrete volume of the designed tissue scaffold
unit cell is represented by a specific feature which is bio-
mimetic to the internal architecture patterns of the re-
placed tissues. We further use the feature patterns and
architectures to design desirable pores, pore sizes
and shapes, and its distribution in the scaffold internal


C 2004 Portland Press Ltd
54 W. Sun and others

Figure 5 CAD-based bone structure model

Table 2 Comparison of 3D reconstructive methods for generating a CAD-based femur model

3D reconstruction method Process File size Quality

Medical CAD Easiest and quickest, but may File size: medium Poor
not be suitable for complex IGES conversion process time: long
tissue geometry IGES: 266 kB
CAD (Pro/ENGINEER
C
): 309 kB

Reverse engineering
A longer process, but suitable
for complex tissue geometry,
since control is achieved at
every level
Initial file sizes in the point form are not Best
high, but final CAD model may involve
comparatively large file sizes
Point: 256 kB (around 7732 points)
IGES: 266 kB (102 NURBS patches)
CAD (Pro/ENGINEER
C
): 298 kB

Stereolithography method
Quick method to arrive at a
CAD model, but not
suitable for complex tissue
geometry; geometry not editable
Initial STL file size maybe high, resulting Average
in large CAD model IGES file size
STL: 1.82 MB (38252 triangles)
IGES: 9.83MB (2316 NURBS patches)
CAD (Pro/ENGINEER
C
): 10.3 MB

structure, so the required mechanical and biological
conditions can be met. We then select the specific feature
primitive as the unit-cell internal architecture according to
the required mechanical or biological functions, and use the
heterogeneous CAD modelling approach to construct unit
cells for bone scaffold, in considering the anatomical and
topological requirements for bone replacement. Samples of
CAD-based cellular-unit-cell-rendered images are shown in
Figure 6. A finite element method (ABAQUS [12]) was then
used to predict the unit-cell effective mechanical properties
as shown in Figure 7 for the unit cell with characteristic
geometrical parameters (Figure 7a), along with its applied
boundary condition (Figure 7b) and the contour plot of the
reaction force (Figure 7c).
The geometry of the unit cell used in the finite element
analysis is defined as a 4.5 mm × 4.5 mm × 4.5 mm block


C 2004 Portland Press Ltd
with four open pores on each face. This cellular-like scaffold
unit cell model was generated by Pro/ENGINEER
C
software,
and converted into IGES format and then imported to
ABAQUS for finite element analysis. The relationship of
the porosity with the overall geometry of the unit cell
for the homogeneous square unit cell with square pores,
is determined by:
N[{L ∗ l 2∗ (f /2)} − {(f /2 − 1) · l 3 · N/2}]
P = (4)
L3
where N is the number of the pores in the unit cell, f is
the total number of faces that contain pores, L is the size of
square unit cell and l is the size of pore (Figure 7a).
A total 8353 four-node tetrahedral elements were used
in the finite element analysis. A schematic illustration of
applying boundary conditions and loading conditions for
 Computer-aided tissue engineering: application to biomimetic modelling and design of tissue scaffolds 55

Figure 6 Samples of the designed scaffold unit cells

Figure 7 Finite element method for predicting unit cell stress and deformation

(a) Unit cell with characteristic geometrical parameters; (b) applied boundary conditions to the unit cell; (c) FEA (finite element analysis) contour plot for stress
distribution within the scaffold.

calculating effective Young’s modulus Exx is shown in Fig- where Ax is the area of cross-section of the face Sx2 and Rx is
ure 7(a). The face given by Sx1 is constrained in the x- the average reaction force on the surface Sx2 . Since the
direction i.e. Ux = 0 and the face given by Sx2 is given unit cell is symmetric in the x, y and z directions, we
a displacement of Ux = 0.001 Lx which is equivalent to obtain Exx = Eyy = Ezz = Eeffective . A summary of the results
ε x = 0.1 % where Lx is the length of the face. This boundary predicted by finite element method for scaffolds with various
condition is intended to obtain a uniform displacement field porosities at 26, 41, 59, 76 and 83 % levels for three
throughout the model in the x-direction, thus enabling us to different biomaterials {hydroxyapatite, PLGA [poly(DL-
calculate the effective E in the x-direction, Exx . The effective lactic-co-glycollic acid)] and PLLA [poly(L-lactic acid)]} is
Young’s modulus of the cell can then be obtained by using presented in Table 3 and further plotted in Figure 8.
the relationship given below: From the QCT characterization of the proximal femur,

Rx  the bone Young’s modulus varies from 0.6 to 2.0 GPa.
σx Rx
Exx = =
USxx 2 =
A
(5) From a biological point of view, we know that a desirable
εx L
0.001 · Ax scaffold structure should have a porosity ranging from
x

C 2004 Portland Press Ltd
56 W. Sun and others

Table 3 Summary of results obtained by FEA

Effective modulus of unit cell at different porosity levels (GPa)

Material Young’s modulus (E ) (GPa) Porosity (%) . . . 26 41 59 76 83

Hydroxyapatite 2.0 1.362 0.9 0.6734 0.328 0.2764

PLLA 2.7 1.7466 1.2564 0.909 0.44 0.375
PLGA 4.1 2.652 1.7466 1.38 0.673 0.567

are anisotropic and can be determined through a composite

lamination theory. At the laminar level, a unit cell is used to
calculate the laminar effective properties. The average linear
stress-strain relations of the unit cell in the kth layer in this
model can be expressed by:
σ i , j = C ij pq εq p (6)
where

n    n
C i(k)3 C 3(k)j C 13(k)
C ij = α k C ij(k) − + α (k)

Figure 8 Effect of unit-cell material and porosity on unit-cell mechanical
properties
L-PGA, PLGA; L-PLA, PLLA.
55 to 70 %. In this regard, the unit cell made of hydroxy-
apatite material with 59 % porosity and effective Young’s
modulus of 0.6734 GPa barely meets both biological and
mechanical requirements. PLLA-based unit cells with about
40–60 % porosity do give a better option as a scaffold
material for the proximal femur. Candidate unit cells can
then be selected from the predicted curves shown in Fig-
ure 8 in our review in this issue [2].
Once the appropriate unit cell has been identified
with the matched porosity, interconnectivity and mechanical
properties, a contour bone structure reconstructed from
CT/MRI images will be filled in with the selected unit-cell
architecture. A constructive heterogeneous solid geometry
algebra [15,16] can be used to combine the unit-cell
architecture and the replaced bone anatomical structure
to achieve the final shape of the replaced bone tissue
scaffold. An example of how a basic unit cell was intersected
within Geomagics software with a bone structure that
needs to be replaced to form a tissue scaffold of the exact
internal architecture and external anatomy is schematically
illustrated in Figure 8.
We can also consider the structural heterogeneity of
bone as an n-layered laminated composite. This model is
analogous to the real bone replacement/scaffold structure
produced through layered solid-freeform-fabrication tech-
nology [for example SLA (STL apparatus), LOM (laminated
objective manufacturing) and 3D printing (3DP)]. The
overall effective mechanical properties of bone structure


C 2004 Portland Press Ltd
k=1
C 33(k) k=1
C 33(k)
 n −1
 n
C 3(l j)  α (k)
× α (l )
(l )
(i = 1,2,3 . . .) (7)
l =1
C 33 k=1 C 33(k)
represents the average effective elastic tensor of the bone,
Cij the elastic components of the kth layer and α k = hk /h, with
h denoting the total height of the bone and hk the height of
each layer.
Freeform fabrication of tissue scaffold
A sample process-planning scenario to design and fabricate
tissue scaffold structures is depicted in Figure 9. In the cus-
tom implant planning stage, the CT/MRI images of the de-
fected region are obtained and reconstructed by the
method as we have already outlined in the previous section.
Through reverse-engineering techniques, a CAD model
of the region of interest is made. At the same time,
mechanical characterization techniques would be employed
to obtain the mechanical properties of the region to be
replaced. On the basis of the characterization process
and biological requirements, an appropriate unit cell is
designed and, along with constructive heterogeneous solid-
geometry techniques, an appropriate scaffold structure
obtained. The CAD file is then converted into a STEP
(standard for exchange of product model data) file format
with designed geometrical and material information [17,18].
Using an appropriate 3D modelling kernel, the scaffold
implant reconstructed from the STEP definitions is then
sliced according to the slice parameters set using the slice
module of the software. The slice module based on the
direct slicing approach would work on the basis of the B-
rep structure in the STEP file [19]. The slicing is achieved
on the basis of the ray-casting method, where the object is
 Computer-aided tissue engineering: application to biomimetic modelling and design of tissue scaffolds
Figure 9 Flowchart of process planning for fabricating tissue scaffold
sliced by a series of parallel planes, each of the planes having
rays cast out across them. The intersection points of the ray
with the object define an entry point and exit point. A set of
these points defines raster lines. The slicing module defines
the slice layers as scan raster lines to support the open-
end architecture that is required for the scaffold structures.
Once these scan lines are defined, complete information that
consists of the slice layers as well as the material information
is stored in a print-job database that serves as a temporary
storage as well as for future retrieval. The scan lines are
converted into machine-code instructions that are then sent
to the machine for fabrication.
A sample of cellular poly(6-hexanolactone) [poly(ε-
caprolactone)] scaffold fabricated by our in-house-deve-
loped micronozzle-freeform-extruding-deposition system is
shown in Figure 7 of [20]. The scanning-electron-micro-
scopic characterization shows that the thus fabricated scaf-
fold micro-architectures could be achieved at about
250 µm scale level with excellent uniformity of the fill gaps,
the depositing struts, and the internal pore connectivity as
shown in Figure 7 of [20]. This extruding deposition process
directly fabricates the designed bone tissue scaffolds ac-
cording to the design specifications defined in the bio/
tissue informatics modelling developed through the biomi-
57
metic design approach, for example, by converting the
biomimetic designed scaffold internal architectures into
layered deposition patterns. This process eliminates mate-
rial preparation and indirect casting in the scaffold fabri-
cation, thus opening opportunities for many biomaterial
applications.
Conclusions
This paper presents an application of the CATE approach for
biomimetic modelling and design of a bone-tissue engineered
structure. The process of the modelling development,
QCT characterization method, tissue representative unit-
cell structural analysis and a CAD-based scaffold bone
design have been described. Results of the effect of the
unit-cell materials and the designed porosities on the unit-
cell mechanical properties for hydroxyapatite, PLGA and
PLLA materials have been given. The CATE approach
presented provides an effective tool for modelling, design
and fabrication of complex bone-tissue scaffolds under
multiple biological and biophysical design constraints. The
calculated E from the characterization is in accordance with
published data for cancellous bone in the range 0.5–1.5 GPa.


C 2004 Portland Press Ltd
58 W. Sun and others
It is noteworthy that both cancellous and cortical bone
has been ‘smeared together’ in our laminated composite
model, hence the slightly higher values of E obtained by
our method. The curves presented in Figure 8 can aid in
the selection of a unit cell with the required porosity and
mechanical properties to match that of the replaced bone.
However, only one Young’s modulus was calculated and
used in the present study. Process-planning methods and the
extruding-deposition process provide a road map on how an
appropriate bone scaffold can be fabricated. The use of STEP
to transfer the CAD model from the design software to
the in-house-fabrication planning software entails advantages
of transferring the complete product model data along
with bone heterogeneity information in the future. While
the value of a macroscopically designed biomimetic model
as opposed to a synthetic model is open to debate, the
truly biomimetic model will assure that elements such as
level of tissue perfusion, mechanical properties and implant
architecture are as compatible as possible with the intended
host.
This application study also shows the capability and
potential of CATE. CAD, in conjunction with advanced
modelling and freeform fabrication technology, can create
scaffolds that go beyond conventional designs currently used
in tissue engineering. It can create complex structures with
functional components that are difficult, if not impossible,
to create with conventional tissue scaffold fabrication
techniques. The application of CATE to tissue engineering
also allows one, for example, to control and design (1) the
overall shape of the scaffold to match a patient’s CT or
MRI data, (2) the internal architecture (pore size, porosity,
interconnectivity) by using the power of CAD techniques
and (3) a heterogeneous scaffold with different mechanical
and geometrical properties, and to freeform fabrication of
the designed scaffolds.
The field of CATE is still in its infancy. There is still much
to be learned and discovered. With advances in computing
power, with new software specifically developed for tissue
engineering and biological systems, and with improvements
in freeform fabrication process, one can envisage that CATE
will play a significant role in the future of tissue engineering.
The bio/tissue informatics model developed on the basis
of the CATE approach will eventually provide precise design
and control of architecture with multi-material printing for
different types of growth factors, cells, adhesion factors and
scaffold materials.
Acknowledgements
This research was partially supported by the National
Science Foundation through grant 0219176, and by Therics
Inc., Princeton, NJ, U.S.A.


C 2004 Portland Press Ltd
References
1 Sun, W. and Lal, P. (2002) Comput. Methods Programs
Biomed. 67, 85–103
2 Sun, W., Darling, A., Starly, B. and Nam, J. (2004) Biotechnol.
Appl. Biochem. 39, 29–47
3 Currey, J. D. (1969) J. Biomech. 2, 477–480
4 Hutmacher, D. W. (2000) Biomaterials 21, 2529–2543
5 Hollister, S. J., Levy, R. A., Chu, T. W., Halloran, J. W., Feinberg,
S. E. (2000) Int. J. Oral Maxillofacial Surg. 29, 67–71
6 Darling, A., Starly, B., Gomez, C., Fang, Z. and Sun, W. (2003)
SPIE/BIOS 2003: Symposium on Optical Technologies to Solve
Problems in Tissue Engineering and Tissue Mechanics, January
25–31, San Jose, CA, vol. 1, abstract 4961B-40
7 Starly, B., Lau, W., Fang, Z. and Sun, W. (2002) in Recent
Developments in Biomedical Engineering (Vossoughi, J., ed.),
pp. 305–306, Medical and Engineering Publishers, Washington,
D.C.
8 Sun, W., Starly, B., Lau, A., Lau, W., Lim, D., Seliktar, R.,
Youssef, A., Bradbury, T. and Gaylo, C. (2002) Proceedings
of the American Society for Composites, 17th Technical Con-
ference, 12–15 October, Purdue University, West Lafayette,
paper #144
9 Materialise (2002) MIMICS User Manual, Materialise, Leuven
10 Raindrop Geomagic (2002) GeoMagic User Manual, Raindrop
Geomagic, Research Triangle Park, NC
11 Parametric Technology Corporation (2002) Pro/ENGINEER
C
User Manual, Parametric Technology Corporation, Waltham,
MA
12 ABAQUS Inc. (2002) ABAQUS User Manual, ABAQUS Inc,
Pawtucket, RI
13 Rice, J. C., Cowin, S. C. and Bowman, J. A. (1998) J. Biomech.
21, 13–16
14 Rho, J. Y., Hobatho, M. C. and Ashman, R. B. (1995) Med. Eng.
Phys. 17, 347–355
15 Sun, W. and Hu, X. (2002) J. Comput. Aided Design 34,
481–488
16 Sun, W. (2000) J. Comput. Sci. Technol. 15, 27–36
17 Dutta, D., Kumar, V., Pratt, M. J. and Sriram, R. (1998)
International Federation for Information Processing (IFIP),
WG5.2/5.3 PROLOMAT Conference ’98, Trento, Italy,
abstract #45, Kluwer Academic Publishers, Amsterdam
18 Dutta, D. (1998) Comput. Eng. Conf., American Society for
Mechanical Engineers, Atlanta, GA, abstract DETC98/CIE-
5698
19 Starly, B., Lau, A., Sun, W., Lau, W., Bradbury, T., Youssef, A.
and Gaylo, C. (2003) Proceedings of the 14th Solid Freeform
Fabrication Symposium, August 4–6, Austin, TX, abstract #11,
in the press
20 Wang, F., Shor, L., Darling, A., Khalil, S., Sun, W., Güçeri, S. and
Lau, A. (2004) Rapid Prototyping J., in the press
Received 1 July 2003/30 September 2003; accepted 14 October 2003
Published as Immediate Publication 14 October 2003, DOI 10.1042/BA20030109