HAL Id: hal-00751516
https://hal.archives-ouvertes.fr/hal-00751516
Submitted on 13 Nov 2012
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-
entic research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diusion de documents
scientiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Three-dimensional terahertz computed tomography of
human bones
Maryelle Bessou, Bruno Chassagne, Jean-Pascal Caumes, Christophe Pradère,
Philippe Maire, Marc Tondusson, Emmanuel Abraham
To cite this version:
Maryelle Bessou, Bruno Chassagne, Jean-Pascal Caumes, Christophe Pradère, Philippe Maire, et al..
Three-dimensional terahertz computed tomography of human bones. Applied optics, Optical Society
of America, 2012, 51 (28), pp.6738-6744. �10.1364/AO.51.006738�. �hal-00751516�
Three-dimensional terahertz computed
tomography of human bones
Maryelle Bessou,
1
Bruno Chassagne,
2
Jean-Pascal Caumes,
3
Christophe Pradère,
3
Philippe Maire,
4
Marc Tondusson,
4
and Emmanuel Abraham
4,
*
1
PACEA, Univ. Bordeaux, CNRS, UMR 5199, F-33405 Talence, France
2
ALPhANOV, Centre Technologique Optique et Lasers, F-33405 Talence, France
3
I2M-TREFLE, Univ. Bordeaux, ENSAM-CNRS, UMR 8508, F-33405 Talence, France
4
LOMA, Univ. Bordeaux, CNRS, UMR 5798, F-33405 Talence, France
*Corresponding author: em.abraham@loma.u‐bordeaux1.fr
Received 24 July 2012; revised 23 August 2012; accepted 25 August 2012;
posted 27 August 2012 (Doc. ID 173175); published 25 September 2012
Three-dimensional terahertz computed tomography has been used to investigate dried human bones
such as a lumbar vertebra, a coxal bone, and a skull, with a direct comparison with standard radiography.
In spite of lower spatial resolution compared with x-ray, terahertz imaging clearly discerns a compact
bone from a spongy one, with strong terahertz absorption as shown by additional terahertz time-domain
transmission spectroscopy. © 2012 Optical Society of America
OCIS codes: 110.6795, 110.6880, 100.6950, 110.7440, 170.6510.
1. Introduction
Radiography is considered as one of the most efficient
techniques to analyze biological tissues such as hu-
man bones. It offers submillimeter spatial resolution
and highly contrasted images depending on the
radiodensity, the thickness, and the composition of
the materials [
1,2]. In the field of three-dimensional
(3D) imaging, x-ray computed tomography (CT) is an
ubiquitous technique, largely employed in medical
imaging, which provides cross-sectional images of
an object by analyzing the radiation transmitted by
the sample through different incidence angles.
However, the use of x-ray is sometimes not recom-
mended regarding the ionizing power of the radiation
and the difficulty of sample radiometric dating after
x-ray irradiation [
3]. Radiography also requires strict
safety rules based on the knowledge of the radiation
effects and on the principles of protection. Only li-
censed users can manipulate x-ray producing devices
and perform radiographic examination, respecting
safe operating procedures and ensuring the wearing
of personal dosimeters. Besides, an under-exploited
domain of the electromagnetic spectrum, known as
the terahertz (THz) spectral region, has recently
emerged as a possible powerful candidate for the ana-
lysis of opaque objects. Being nondestructive and con-
tactless in nature, the THz wave can penetrate into
nonconductive and nonpolar materials and offers
complementary spectroscopic data for a better diag-
nosis and understanding of materials. THz imaging
has been widely employed in the field of homeland
security and more recently for the investigation of
artwork related to cultural heritage in collaboration
with curators from museums [
4–7].
Surprisingly, up to now, very few studies have been
conducted concerning the analysis of human bones by
THz radiation. Mechanical properties of compact
bone have been first investigated by THz time-
domain transmission spectroscopy in the 0.1–
1.25 THz spectral range [
8]. No correlation was found
between the values of the Young’s modulus and
the spectroscopic THz parameters. The authors
1559-128X/12/286738-07$15.00/0
© 2012 Optical Society of America
6738 APPLIED OPTICS / Vol. 51, No. 28 / 1 October 2012
concluded that THz spectroscopy cannot rival with
dual energy x-ray absorptiometry (DEXA) to identify
the bone mineral density. THz pulsed imaging has
been used to measure the early symptoms of osteoar-
thritis [
9]. In this study, THz reflection images of
excised rabbit femoral condyles demonstrated that
multiple reflections can provide quantitative infor-
mation on cartilage substructure. Baughman et al.
presented a spectral imaging of osseous tissues
through time-domain transmission spectroscopy.
However, their study was limited to identifying re-
gions of different osseous tissue type within a thin
cross section of the distal epiphysis from a chicken
bone [
10]. At last, Öhrström et al. compared x-ray
and pulsed THz imaging in the case of an artificially
embalmed ancient Egyptian human mummy hand
and a macerated human lumbar vertebra [
11]. They
pointed out that conventional radiography naturally
provided higher spatial resolution, whereas THz
imaging was well-suited for the investigation of soft
tissues such as cartilaginous structures.
None of these studies presented the alternative po-
tential of THz radiation for 3D imaging using CT.
Since the first demonstration of THz CT by Ferguson
et al. in 2002, few papers have been actually published
since the technique suffers from strong limitations
such as beam diffraction and severe Fresnel losses
for high refractive index samples [
12–18]. In this
paper, after a characterization of the THz spectro-
scopic properties of human bone, we present the first
3D THz CT analysis of selected dried human bones
associated with a direct comparison with conven-
tional radiography. This study is focused on dried hu-
man bones, owing to the strong absorption of THz
radiation by water. Therefore, the main goal of our
work is not directly addressed to medical applications
but rather to archeology or biological anthropology.
2. Experimental Methods
A. Materials
Human bones consist of compact and spongy bones.
Compact bone is more dense, smooth, and homoge-
neous, whereas spongy bone is an open network
structure, with small and irregular adjacent cavities.
As the sample hydration state is of primary impor-
tance for THz transmission, our study has been
limited to dried samples. For spectroscopic studies,
pellets of bone samples have been prepared from a
dried human femur. The compact bone pellet is
very thin (0.265 mm) with a thickness precision of
0.005 mm obtained with a digital calliper. Such a
thin pellet was required in order to properly measure
the amplitude of the THz transmitted wave, as ex-
plained in the following section. The spongy bone
pellet is much thicker (2 mm) owing the fragile nat-
ure of the highly scattering internal structure.
Consequently, the spongy bone pellet will highly
scatter the incoming THz radiation.
For imaging, we selected various relevant human
bones such as a lumbar vertebra, a skull, and a coxal
bone. The human bones were obtained from osteolo-
gical collections at de la Préhistoire à l’Actuel: Cul-
ture, Environnement et Anthropologie (PACEA).
They were selected according to their morphology
and their internal structure. In this current study,
only anatomic features useful for the optimal under-
standing of the images are indicated. The vertebrae
belong to the irregular bones. They consist of a body
and a vertebral arch constituted with costal (later-
ally) and spinous processes (posteriorly). A lumbar
vertebra was chosen because of its morphology, with
a broader an d thicker body compared with other ver-
tebrae [Fig.
3(a)]. These morphometric features will
allow us to optimize the reading of the images. The
processes are thin and mainly made of compact bone,
whereas the vertebral body is mostly constituted of
cancellous bone enclosed with a thin layer of compact
bone. The vault of skull and the coxal bones consti-
tute flat bones. Thin , flattened, and usually a bit
curved, they have two roughly parallel compact bone
surfaces, with a layer of spongy bone between. For
the skull, only the vault is composed with flat bones,
whereas the face and base ones are classified as irre-
gular bones. Numerous flat and thin bones with var-
ied morphology compose the face, and their specific
architecture creates several cavities (frontal sinuses,
maxillary sinuses, nasal cavity, etc.). We will discuss
in the Section 3 the influence of these cavities on THz
imaging. In our study, the skull is interesting be-
cause of its important volume, the presence of the
two types of bone tissues (spongy bone and compact
bone), and the numerous adjoining empty spaces of
the face. The current analyzed skull was formerly
partially restored with plaster particularly both
in the lateral and superior parts of parietal bones
[Fig.
4(a)]. Finally, we propose to analyze a coxal bone
that is a great flat bone [Fig.
5(a)]. The ilium wing
does illustrate the specific feature of flat bones with
spongy bone sandwiched between two thin layers of
compact bone. This bone presents differences of
thickness both in its superior (ilium) and inferior
(pubis and ischium) parts. X-rays can distinguish
very well these differences. We will see in the
Section
3 how THz radiation can differentiate them.
B. Methods
THz time-domain transmission spectroscopy (TDS)
was performed in order to measure the THz trans-
mittance of human bone. To this aim, we used an am-
plified laser system (800 nm, 1 kHz, 1 mJ, 50 fs) to
produce a THz beam consisting in a single-cycle THz
wave generated by the interaction of the fundamen-
tal and the second-harmonic laser pulses in ionized
air [19]. After transmission through the sample, this
wave is measured by electro-optical sampling in a
100 μm-thick h110i GaP crystal in the 0.3–8THz
spectral range, the lower frequency being limited
by the THz emission within the plasma and the
higher one by phonon absorption in the GaP crystal.
This broadband spectrometer makes possible the
evaluations of the frequency-dependent refractive
1 October 2012 / Vol. 51, No. 28 / APPLIED OPTICS 6739
index and absorption coefficient of the samples. The
determination of the refractive index is important in
order to quantify the deviation of the THz beam in
the case of oblique incidence due to Fresnel laws,
whereas the characterization of the absorption coef-
ficient will indicate the global attenuation of THz
radiation after propagating through bone samples.
The experimental setup of the continuous wave
THz imaging system—shortly called the “THz
scanner”—has already been described [
20]. Briefly, it
consists of a compact millimeter wave Gunn diode
(110 GHz, 20 mW) coupled with a horn antenna
(Fig.
1). The output beam is focused on the sample,
which is positioned on two-axes XY motorized
stages. At the sample position, the beam diameter is
diffraction limited and the final image is obtained
point-by-point by raster scanning the sample in both
horizontal and vertical directions. For the detection,
we used a commercial low-cost pyroelectric sensor
(Gentec EO) and an optical chopper connected to a
lock-in amplifier. A two-dimensional (2D) transmis-
sion image of the sample is obtained by moving
the object in the X and Y directions with a scan step
of 1 or 2 mm in both directions. Owing to the low fre-
quency of the emitting source, the spatial resolution
of the THz scanner is limited to a few millimeters
(110 GHz corresponds to a wavelength of 2.7 mm).
This is a strong limitation of the system for the ana-
lysis of human bones that require for imaging a spa-
tial resolution better than 1 mm. With a scan speed
of 5 pixels ∕s, the acquisition time for a (100 × 100 )
pixels image size is about 30 min. This portable
THz scanner is well adapted for on-site imaging since
its volume is about (500 mm × 500 mm × 500 mm)
with a total weight less than 20 kg. Moreover, as
pointed out in the introduction, the imaging system
is easy to install and completely safe for the sample
and the operator, unlike radiography.
For 3D reconstruction, the sample is rotated in or-
der to provide a different visualization of the object,
such as in x-ray CT. From these tilted series, we are
able to construct the sinogram of the object that re-
presents, for a given horizontal slice, the evolution of
the transmitted THz amplitude as a function of the
rotation angle. Here, to get a reasonable acquisition
time for 3D imaging, we selected a rotation step of
10°. In this case, we obtained the correspo nding 18
projections in nearly 9 h. In this procedure, it is
important to position the rotation axis exactly in
the middle of the 2D transmission images for any
projection angle to avoid any future reconstruction
artifacts due to lateral shifting. Moreover, as ex-
plained in our previous work, the 10° rotation incre-
ment has been selected as a compromise between
acquisition time and 3D imaging quality represented
by spatial resolution, contrast, intensity, and geo-
metric preservation [
26]. For the final 3D reconstruc-
tion, we used the stand ard backprojection of the
filtered projections (BFP) algorithm [
21]. This recon-
struction process is based on the inverse radon trans-
form [
22], which computes the final pixel values from
the filtered projections. It is widely developed in
x-ray CT scan imaging and also used in THz CT.
As will be shown in the following section, this stan-
dard algorithm is suitable for the final 3D recon-
struction even if it is well known that it suffers
from several disadvantages such as beam hardening,
noise sensitivity, and geometric degradation in case
of an insufficient number of projections. In a previous
study, we tested alternative iterative reconstruction
methods for THZ CT, such as the simultaneous alge-
braic reconstruction technique (SART) [
23] or the or-
dered subsets expectation maximization (OSEM)
method [
24,25], with equivalent accuracy and ima-
ging quality to BFP [
26]. For a limited number of
projections (typically less than 15), we emphasized
a quantitative degradation of the BFP reconstruc-
tion, whereas the SART method can already offer op-
timized reconstruction quality. However, in this
study, we selected the BFP algorithm for 3D recon-
struction since it is extensively proposed in most of
CT sof tware tools.
For standard radiography, we used a medical x-ray
tube including a rotating anode with a tungsten–
rhenium target on a molybdenum core. In order to
properly visualize the structural characteristics of
the human bones, three physical para meters were
selected: the potential difference between the anode
and the cathode of the x-ray tube corresponding to
the penetrating power (42–45 kV), the tube current
(50 mA), and the beam intensity related to the expo-
sure time (25 mAs).
Fig. 1. (Color online) (a) Experimental setup. C, optical chopper; M, off-axis parabolic mirror (f
0
150 mm); L, Teflon lens (f
0
60 mm);
S, sample; D, pyroelectric detector. (b) Photograph of the THz scanner analyzing a human lumbar vertebra.
6740 APPLIED OPTICS / Vol. 51, No. 28 / 1 October 2012
3. Results and Discussion
A. THz Time-Domain Transmission Spectroscopy
At first, the spongy bone pellet was analyzed but the
amplitude of the THz transmission was too weak
compared to the noise level to quantitatively analyze
this sample. The reason arises from the highly
diffusive nature of this sample even if its thickness
has been decreased as much as possible, as pre-
viously explained. For a quantitative study of this
sample, a more precise analysis is in progress, taking
into account the real scattering nature of the sample,
according to the recent work of Joly et al. [
27].
THz time-domain trans mission spectroscopic
properties of the compact bone pellet is presented
in Fig.
2. In spite of the broadband time-domain spec-
trometer (0.3–8 THz), the data are limited to the
0.3–2.75 THz spectral range owing to the strong
absorption of THz radiation at higher frequencies.
We observed that the refractive index slightly in-
creases from 1.92 to 1.97 within the 0.3–1.9 THz
spectral range (blue curve). This measurement
slightly differs from that of Stringer et al. (mean re-
fractive index between 2.2 and 2.6, derived from the
time-domain signal) [
8]. One reason can be the pro-
blem of indirect comparison of biological tissues,
which is always difficult owing to the inevitable pre-
sence of inhomogeneous structures within the sam-
ples. At higher frequencies, the amplitude of the
refractive index fluctuates owing to poor signal-to-
noise ratio. For THz CT, an important conclusion of
this measurement is that the relatively high refrac-
tive index of compact bone will significantly devi-
ate the direction of propagation of THz radiation in
the case of oblique incidence onto the sample, accord-
ing to the Fresnel laws. The absorption coefficient of
compact bone also gradually increases from 10 to
420 cm
−1
in the 0.3–2.75 THz spectral range (red
curve). This additional important result shows that
THz imaging of human bone is almost impossible
at frequencie s higher than 1 THz owin g to a very
high THz absorption, probably due to the presence
of inorganic materials into the samples, such as
minerals (calcium) that represent the major compo-
nents of the osseous tissues. To minimize the attenua-
tion of THz radiation, it is more suitable to select a
sub-THz frequency in spite of the degradation of
imaging transverse resolution. The extrapolation of
previous measurements provides an absorption coef-
ficient less than 5 cm
−1
at 110 GHz, corresponding to
the emission frequency of the THz scanner. This esti-
mation is in agreement with the direct evaluation of
the absorption coefficient of compact bone α 2.5
0.5cm
−1
at this frequency obtained with use of
the 110 GHz source. Considering the signal-to-noise
ratio of 10; 000∶1 for the THz scanner, we can con-
clude that the total thickness of the compact bone
should be roughly limited to 15 mm for effective ima-
ging. As pointed out previously, besides the limited
spatial resolution, this weak penetration depth con-
stitutes another strong limitation of THz imaging
compared to x-ray for the analysis of human bones.
B. 2D THz Computed Tomography and Radiography
Figure
3(a) shows the photograph of the dried human
lumbar vertebra (superior view). The radiograph ea-
sily distinguishes with a submillimeter spatial reso-
lution the two different types of bone structure, in
particular the spongy bone microstructure [Fig.
3(b)].
As previously mentioned, the body of the vertebra is
mainly composed with spongy bone, appearing in
black on the radiograph and reflecting its low radio-
pacity. On the contrary, the limits of the body and the
vertebral arch are made of more radiopaque compact
bone. In spite of the reduced spatial resolution, THz
imaging also reveals the density varia tions of the
sample, indicating that compact bone exhibits higher
THz absorption than the spongy one at 110 GHz
[Fig.
3(c)]. This additional observation was not possi-
ble by previous THz time-domain spectroscopy, as
described above. We can also notice that the micro-
spaces into the spongy bone do not disturb the pro-
pagation of the sub-THz wave [black area in the body
of the vertebra, Fig.
3(c)].
Similarly, Fig.
4(a) shows the photograph of the
dried human skull (right lateral view). The radio-
graph of the skull exhibits the limits of the cranial
vault, the coronal suture indicated by the red arrow
in the figure, the whole details of the face with the
different cavities and the maxilla [Fig.
4(b)]. The
dark zones on the radiograph arise from the empty
spaces. THz imaging provides some similarities such
as the important THz beam absorption by the vault
bones and regions where the plaster is present (in
the lateral and superior parts of parietal bones)
[Fig.
4(c)]. The base and the face of the skull exhibit
the same THz absorption except at the level of the
orbital cavities. This can be explained by the multi-
ple spaces in the base and the face of the skull that
scatter the THz wave. Besides, in the case of a single
cavity with smooth walls like those of the cranial
vault, we clearly observed that the THz beam is
not disturbed providing a higher THz transmission.
Finally, the imaging of the coronal suture by both
0,0 0,5 1,0 1,5 2,0 2,5
1,80
1,85
1,90
1,95
2,00
2,05
2,10
Refractive index
Frequency (THz)
0
50
100
150
200
250
300
350
400
450
Absorption coefficient (cm
-1
)
Fig. 2. (Color online) Refractive index (blue curve) and absorption
coefficient (red curve) of compact bone.
1 October 2012 / Vol. 51, No. 28 / APPLIED OPTICS 6741