Journal of Microscopy, Vol. 00, Issue 0 2016, pp. 1–9 doi: 10.1111/jmi.

12507
Received 17 June 2016; accepted 30 October 2016

Micro-CT versus synchrotron radiation phase contrast imaging

of human cochlea
M . E L F A R N A W A N Y †, S . R I Y A H I A L A M †, S . A . R O H A N I ‡, N . Z H U §, S . K . A G R A W A L †, ‡, , #, ∗ &
H . M . L A D A K †, ‡, , #, ∗
†Department of Otolaryngology-Head and Neck Surgery, Western University, London, Ontario, Canada
‡Biomedical Engineering Graduate Program, Western University, London, Ontario, Canada
§Bio-Medical Imaging and Therapy Facility, Canadian Light Source Inc., University of Saskatchewan, Saskatoon, Saskatchewan, Canada
Department of Medical Biophysics, Western University, London, Ontario, Canada
#Department of Electrical and Computer Engineering, Western University, London, Ontario, Canada

Key words. Human cochlea, image segmentation, microcomputed

tomography, phase contrast imaging, synchrotron radiation,
three-dimensional model.

Summary hearing loss. One component of the CI is an electronic im-

plantable electrode array that is inserted into the cochlea to
High-resolution images of the cochlea are used to develop
stimulate the auditory nerve and restore a sense of sound to
atlases to extract anatomical features from low-resolution
the patient. The cochlea is divided into two main fluid-filled
clinical computed tomography (CT) images. We compare vi-
compartments: the scala tympani (ST) and the scala vestibuli
sualization and contrast of conventional absorption-based
(SV). A smaller fluid-filled duct, the scala media, separates the
micro-CT to synchrotron radiation phase contrast imaging
ST from the SV. The boundary between the SV and scala media
(SR-PCI) images of whole unstained, nondecalcified human
is formed by the very thin Reissner’s membrane (RM), and the
cochleae. Three cadaveric cochleae were imaged using SR-PCI
boundary between the ST and the scala media is formed by
and micro-CT. Images were visually compared and contrast-
the basilar membrane (BM). The array should preferentially
to-noise ratios (CNRs) were computed from n = 27 regions-
be inserted into the ST instead of the SV to reduce the risk
of-interest (enclosing soft tissue) for quantitative comparisons.
of intracochlear trauma (Zrunek et al., 1980; Adunka et al.,
Three-dimensional (3D) models of cochlear internal structures
2005; Adunka & Buchman, 2007). Moreover, deeper inser-
were constructed from SR-PCI images using a semiautomatic
tion towards the cochlear apex provides better low-frequency
segmentation method. SR-PCI images provided superior vi-
stimulation and speech understanding (Hamzavi & Arnoldner,
sualization of soft tissue microstructures over conventional
2006); however, intracochlear compartments become quite
micro-CT images. CNR improved from 7.5 ± 2.5 in micro-CT
narrow at the apex, and this area can be difficult to discern on
images to 18.0 ± 4.3 in SR-PCI images (p < 0.0001). The semi-
imaging.
automatic segmentations yielded accurate reconstructions of
Although various modalities have been used for imaging the
3D models of the intracochlear anatomy. The improved visu-
human cochlea, computed tomography (CT) is currently the
alization, contrast and modelling achieved using SR-PCI im-
standard imaging modality for postoperative CI surgery eval-
ages are very promising for developing atlas-based segmenta-
uation because it provides three-dimensional (3D) positional
tion methods for postoperative evaluation of cochlear implant
information and offers excellent visualization of bone in the
surgery.
presence of the CI (Skinner et al., 1994). However, clinical CT
does not provide high enough spatial resolution and contrast
Introduction to visualize the BM.
Micro-CT, with its higher spatial resolution, has been used
Cochlear implant (CI) surgery is a procedure used to restore as a visualization tool for evaluation of the surgical aspects
hearing in patients with severe-to-profound sensorineural of newly developed CI electrodes such as electrode position
relative to the BM and possible damage to the BM (Postnov
∗
Co-senior authors. et al., 2006; Teymouri et al., 2011). In order to derive
Correspondence to: Mai Elfarnawany, Department of Otolaryngology, Western Uni-
patient-specific anatomical boundaries from clinical CT im-
versity, 1151 Richmond Street, London, Ontario, Canada N6A 3K7. Tel: +1-519-
661-2111, ext. 86551; fax: +1-519-661-2123; e-mail: melfarna@uwo.ca
ages, image analysis methods have been developed in which

C 2016 The Authors

Journal of Microscopy 
C 2016 Royal Microscopical Society
2 M. ELFARNAWANY ET AL.
a high-resolution statistical atlas describing average cochlear
anatomy and its variations is derived from micro-CT images
and is used to segment (i.e. partition) the BM, ST and SV
from lower resolution clinical CT images (Noble et al., 2013;
Kjer, 2016). Note that because the RM is often not visible in
published micro-CT images, the scala media cannot be differ-
entiated from the SV, and both ducts are simply segmented as
one unit that is collectively referred to as the SV (Noble et al.,
2013; Kjer, 2016).
Micro-CT images still suffer from problems such as artefacts,
noise, low contrast resolution and restriction on specimen
size (Shibata et al. 2009). These problems limit the ability
to visualize the BM along its entire length from the cochlear
base to apex, which makes definitive differentiation of the ST
from the SV difficult without using histological sections as
a reference (Teymouri et al., 2011; Bellos et al., 2014). This
inability to identify the BM makes segmentation of the BM and
of the ST and SV labor-intensive and time-inefficient because
of the need for manual intervention when developing digital
geometric models of these structures for atlas construction
(Skinner et al., 2007; Braun et al., 2012).
The use of synchrotron radiation (SR)-based X-ray imag-
ing can potentially overcome these shortcomings. SR provides
coherent collimated X-rays with high photon flux and highly
stable sources. These high-energy X-ray beams pass through
dense structures easily, making inner structures visible. SR-
based X-ray absorption imaging has been used to image hu-
man cochleae, but staining and dehydration of the specimens
was required to improve the visualization of soft tissues (Müller
et al., 2006; Lareida et al., 2009). It has been reported that
staining specimens causes distortion and shrinkage of the BM
due to dehydration (Brown et al., 2002).
The requirement for staining and the challenge of dehydra-
tion can potentially be avoided with the use of phase contrast
imaging (PCI). Unlike conventional absorption-based imag-
ing, PCI is a technique that depends on the phase shift of
an X-ray beam as it propagates through materials of differ-
ent X-ray refractive index and transforms it into measur-
able amplitude information. The additional phase shift is ob-
served at the edges of structures, which enhances the contrast
where the difference in X-ray absorption is usually weak (Xu
et al., 2001; Bartels et al., 2013). By combining the high pho-
ton energy hard X-rays produced using SR with PCI (called
‘SR-PCI’ henceforth), weakly X-ray absorbing microstruc-
tures are visible in the presence of the dense bone encapsu-
lating the cochlea. With its strong visualization capability,
this method does not require dehydration or staining of the
tissue.
SR-PCI has been previously used to visualize cochlear mi-
crostructures in animals (Shu et al., 2007; Richter et al., 2009;
Rau et al., 2012) but has not been used on human cochlear
specimens to date. In a number of studies by Rau et al., im-
ages depicting microstructures such as the ST, SV, BM, RM,
spiral ganglion (SG), tectorial membrane (TM), organ of Corti
(OC), Spiral Limbus and cochlear nerve (CN) were presented
(Rau et al., 2007; Rau et al., 2012). These features were iden-
tifiable in images acquired at a very high spatial resolution
of 1.22 µm. The imaged microstructures were manually seg-
mented and used to provide a realistic reconstruction of the
mouse cochlea. Nevertheless, this work required excising, fix-
ing and partially decalcifying the cochlea to reduce the X-ray
absorption by the cochlear wall and to ease the detection of
the cochlear soft tissue structures. In a similar study, Shu et al.
were able to visualize the ST, SV and BM of a guinea pig cochlea
using SR-PCI (Shu et al., 2007). They presented images with
10.9 µm spatial resolution, which were used to construct a
simple 3D reconstruction but no segmentation or modelling
of the identified microstructures was done. Similarly, Richter
et al. used a hard X-ray phase contrast source to visualize
the BM, OC and Rosenthal’s canal as well as the SG nerves
of mouse cochleae (Richter et al., 2009). Nonetheless, their
imaging was performed on sectioned slices (20 µm thickness)
of the cochlea which required an intensive sample preparation
protocol and was reported to possibly change the morphology
of the specimen (Brunschwig & Salt, 1997; Edge et al., 1998).
In addition, the intracochlear microstructures were only iden-
tified in two-dimensional reconstructed images of a very high
spatial resolution (0.46 µm) but were not used to construct a
3D model of the cochlea.
In this paper, we present the first use of SR-PCI to visualize
and model the internal microstructures of whole unstained,
nondecalcified human cochleae. Our primary objective is to
demonstrate that SR-PCI can be used to visualize the BM
from cochlear base to apex and to quantitatively compare
its improved contrast to conventional absorption micro-CT.
A secondary objective is to demonstrate a semiautomatic ap-
proach that can be used to segment and model intracochlear
anatomy, particularly of the BM, the ST and SV. Visualization
and modelling the BM, SV and ST is necessary for developing
an atlas-based approach for postoperative evaluation of the CI
placement using clinical CT images.
Materials and method
Sample preparation
Three fresh-frozen adult cadaveric temporal bones were used
in this study. The cadaveric temporal bones were donated to
Western University and permission was granted for use for
the purposes of medical education and research. Following
thawing, a cylindrical cutter having a 40 mm diameter and
60 mm length was used to harvest a core sample of each
temporal bone enclosing the cochlea to match the sample-
holder size of a conventional micro-CT scanner.
The samples were fixed in a 4F1G (3.7% formaldehyde +
1% gluteraldehyde in phosphate buffer) bath for 5 days to
avoid decomposition. 4F1G was used as fixative in this study
because of the combined advantages of faster penetration of
C 2016 The Authors
Journal of Microscopy 
C 2016 Royal Microscopical Society, 00, 1–9
 MICRO-CT Vs. SYNCHROTRON RADIATION PHASE CONTRAST IMAGING
formaldehyde and rapid fixing action of gluteraldehyde
(Metscher, 2009). The samples were rinsed twice and were
dehydrated using an ethanol series consisting of eight stages
(50%, 60%, 70%, 80%, 90%, 95%, 100% and 100%). At
each stage, the samples were immersed in 300 mL of solu-
tion for 30 min. Although sample dehydration was not nec-
essary when acquiring SR-PCI images, it was performed as
commonly reported in the literature to improve soft tissue
contrast in absorption-based micro-CT images (Lareida et al.,
2009; Teymouri et al., 2011; Buytaert et al., 2013). No addi-
tiotnal processing (i.e. staining, sectioning or decalcification)
was performed on the samples.
Experimental setup and image acquisition
SR-PCI imaging. The PCI technique used in this study is the
propagation-based PCI method, which is also known as in-line
PCI. In-line PCI is similar to a typical radiography with an X-
ray source, a sample and a detector but requires the detector
to be placed at a distance from the sample to allow the phase-
shifted beam to interfere with the original beam and produce
measurable fringes. Unlike other PCI techniques (e.g. diffrac-
tion enhanced imaging, crystal or grating interferometry), it
does not require any additional optical elements to be placed
between the sample and detector. The simple setup and the
low stability requirements made it our method of choice for
this study.
SR-PCI combined with CT technique was used for this study.
To acquire the SR-PCI images, each sample was scanned us-
ing the Bio-Medical Imaging and Therapy (BMIT) 05ID-2
beamline at the Canadian Light Source Inc. in Saskatoon,
SK, Canada. The BMIT 05ID-2 beamline is an SR beam pro-
duced by a superconducting wiggler source (Wysokinski et al.,
2015). The beam is filtered using a monochromator and yields
an energy bandwidth of E/E = 10−3 over an energy range
of 20–150 keV. The beam size is 220 × 11 mm at a distance of
55 m from the source and the monochromatic flux is 3 × 10¹²
photons at 20 keV. The imaging setup installed at the beam-
line length of 55 m from source consists of a sample stage and a
charge-coupled-device-based detector system both placed on
a vibration isolation table. The distance between the sample
and detector was set to 2 m and the beam energy was set to
47 keV. Motorized alignment stages were used to align the
sample and detector for high-resolution tomography. The de-
tector (C9300-124, Hamamatsu Photonics, Shizuoka, Japan)
has a 16-bit resolution, 36 × 24 mm field of view and an
effective pixel size of 9 × 9 µm2 . The imaging field of view
was set to 4000 × 950 pixels corresponding to 36.0 ×
8.6 mm and 3000 projections in 180° rotation were acquired
per view. The total acquisition time to capture all projections
was 30 min. Both flat-field and dark-field corrections (i.e.
removal of beam-specific and detector-specific artefacts) were
applied during reconstructions. The reconstructed grayscale
images were of 16-bit resolution.
C 2016 The Authors
Journal of Microscopy 
C 2016 Royal Microscopical Society, 00, 1–9
3
Conventional micro-CT imaging. For comparison, conven-
tional absorption-based micro-CT imaging was performed. To
acquire conventional micro-CT images, the sample was im-
aged using an eXplore Locus micro-CT scanner (GE Health-
care Biosciences, London, Ontario, Canada) at the Robarts
Research Institute, Western University in London. The scan-
ner operated with an x-ray source of 80 kV voltage and a
current of 0.45 mA. Using an angular increment of 0.4°, 900
projections were acquired. For each projection, the object was
exposed for about 4.5 s per frame and an average of five frames
was taken to increase the signal-to-noise ratio. The total ac-
quisition time to capture all projections was 7 h. A modified
cone-beam algorithm (Feldkamp et al., 1984) was used to re-
construct the data into a 3D image volume with an isotropic
voxel size of 20 µm. A subvolume of 200 slices was selected to
span the entire cochlea.
SR-PCI and conventional micro-CT comparison
The SR-PCI and conventional micro-CT images were coreg-
istered into a standard clinically relevant coordinate system
described in the literature (Verbist et al., 2009) to allow for
visual and quantitative comparisons. Fiducial-based registra-
tion using three landmarks (round window, oval window and
apex) was performed in 3D Slicer 4.5 platform (Fedorov et al.,
2012). Following coregistration, matching SR-PCI and micro-
CT images were exported after normalizing their intensities to
0–255 grey-scale levels. The intensity normalization was per-
formed by scaling the full dynamic range of each image while
avoiding clipping or saturation of image intensities.
The contrast of the BM was qualitatively evaluated by plot-
ting intensity profiles of a vertical line marked perpendicular
to the BM in the centre of a 25 × 25 pixel region-of-interest
(ROI) enclosing the BM. The changes in grey-scale intensity
along that line corresponded to variation between the areas
adjacent to and along the thickness of the BM.
The contrast of the BM was quantitatively evaluated by
computing contrast-to-noise ratios (CNRs). The CNR of the
BM was calculated using the following formulae (Bushberg
et al., 2003):
I BM − Ispace
C NR = ,
σspace
where I BM is the maximum grey-scale value in the 25 ×
25 pixel ROI enclosing the BM and Ispace is the minimum
grey-scale value within an ROI of equal size (25 × 25 pixels)
enclosing an entirely air-filled area (i.e. part of the cochlear
ducts adjacent to the BM). σspace is the standard deviation of
grey-scale values within the air-filled ROI. The ROIs that were
used for calculating I BM , Ispace and σspace were outlined in
slices clearly depicting the cochlear partition in the basal turn
(three slices), the middle turn (three slices) and the apical turn
(three slices) giving a total of nine slices per sample. Match-
ing ROIs were used for the previously coregistered SR-PCI
4 M. ELFARNAWANY ET AL.

Fig. 1. Improved intracochlear soft tissue visualization in SR-PCI images over conventional absorption-based micro-CT images. (A), (C) and (E) are
SR-PCI images showing midmodiolar 2D slices of three different cochleae and (B), (D) and (F) are the corresponding conventional micro-CT images.
SR-PCI images show a complete separation between the scala tympani (ST) and scala vestiubli (SV) by the basilar membrane (BM) in the basal, middle
and apical turns. SR-PCI images also show a more detailed view of the spiral ligament (SL), stria vascularis (StV), cochlear nerve (CN) fibres and the spiral
ganglion (SG). Reissner’s membrane (RM) was also visible in (C).

and micro-CT images. The segmentations and calculations
were performed in an open-source platform for biological-
image analysis, Fiji (www.imagej.net; Schindelin et al.,
2012).
Paired Student’s t-tests were used to compare the CNRs of
the two types of images for each of the different cochlear turns.
For each test, n = 9 ROIs (three slices/sample/turn × three
samples) were used. In addition, an overall CNR grouping all
of the turns together was calculated for each type of image and
compared using a paired t-test with all n = 27 images (nine
slices/sample × three samples). In this work, p < 0.05 was
considered to indicate a significant difference.
Segmentation and model reconstruction
A semiautomatic segmentation algorithm, ‘3D GrowCut’
(Vezhnevets & Konouchine, 2005), was applied to extract
models of intracochlear structures necessary for atlas con-
struction. In particular, the ST, SV, BM, spiral ligament (SL)
and stria vascularis (StV) were modelled. Although only the
ST, SV and BM are generally considered in atlases intended for
clinical applications (Noble et al., 2011; Kjer, 2016), the SL
and StV were also extracted in order to highlight the powerful
visualization capabilities of SR-PCI. The GrowCut algorithm
uses manually selected pixels as seeds for structures of interest
and for background. These seeds are iteratively ‘grown’ to be
classified into pixels corresponding to the structures of inter-
est and to background. A fast implementation of the GrowCut
algorithm in 3D Slicer 4.5 platform called ‘FastGrowCut’ (Zhu
et al., 2014) was used for segmentations. The whole FastGrow-
Cut segmentation process was performed on a 3.2 GHz core
i7 PC with 64 GB RAM and GeForce GTX 970 graphics board
with 4 GB RAM with a total running time of 30 min for all
the cochlear structures.

C 2016 The Authors

Journal of Microscopy 
C 2016 Royal Microscopical Society, 00, 1–9
 MICRO-CT Vs. SYNCHROTRON RADIATION PHASE CONTRAST IMAGING 5

Fig. 2. Improved visualization of extracochlear structures in SR-PCI images over conventional micro-CT images. The cochlear nerve (CN) and its
branching fibres are clearly visible in the SR-PCI image (A), but they are very faint in the corresponding conventional micro-CT image (B). The annular
ligament (marked by the circle) connecting the oval window to the stapes footplate is better visualized in the SR-PCI image (C) than in the corresponding
conventional micro-CT image (D).

Results conventional micro-CT image (Fig. 3b), it is hardly visually

Improved soft tissue visualization
The improved visualization of intracochlear soft tissue us-
ing SR-PCI over conventional micro-CT is demonstrated by
Figure 1. The edge enhancement provided by SR-PCI is al-
lowed for visualization of the BM. This provided a clear sepa-
ration between the two main cochlear ducts, the ST and SV.
The BM is not as clearly visualized in conventional micro-
CT, and is missing in some areas closer to the apex. In ad-
dition, the improved contrast in SR-PCI images provided a
sharper and clearer depiction of the SL, StV, CN fibres and SG.
RM is visible in one of the three samples acquired using SR-
PCI.
In general, SR-PCI images also show finer and sharper de-
tails for other extracochlear structures as shown in Figure 2.
For example, the CN and its branching fibres are only visible
in SR-PCI images (Fig. 2a), not conventional micro-CT images
(Fig. 2b). Another example is that visualization of the annular
ligament connecting the oval window and the stapes footplate
is highly improved in SR-PCI images (Fig. 2c) compared to
conventional images (Fig. 2d).
Improved soft tissue contrast
Soft tissue contrast in SR-PCI images is improved both vi-
sually and quantitatively relative to conventional micro-CT
images as shown in Figure 3. The BM appears as a bright line
within the ROI in the SR-PCI image (Fig. 3a), whereas in the
distinguishable within the ROI. The intensity profile through
a vertical line at the centre of the ROI (Fig. 3c) illustrates the
larger signal variation between the background and BM in the
SR-PCI image over the conventional micro-CT image.
CNR values measured from SR-PCI images at the basal,
middle and apical turns were 19.0 ± 5.0, 18.1 ± 4.2 and
17.9 ± 4.2, respectively, whereas the corresponding values
from conventional micro-CT images were 8.1 ± 2.1, 8.1 ±
3.3 and 6.4 ± 1.8. The differences in CNR between the two
imaging modalities were statistically significant (p < 0.001
for all three locations) as summarized in Figure 4. The overall
CNR estimates grouping together all three turns (27 slices)
were 18.0 ± 4.3 and 7.5±2.5 for SR-PCI and conventional
micro-CT, respectively. The overall improvement of CNR in
SR-PCI images was statistically significant (p < 0.0001).
Segmentation and 3D reconstruction
The high contrast and spatial resolution of SR-PCI images al-
lowed very robust and sharp segmentation of the intracochlear
ducts and membranes using the semiautomatic 3D FastGrow-
Cut algorithm. Figure 5 illustrates a sample of the segmenta-
tion results and the corresponding 3D reconstructed model.
Clear separation between SV, ST and the region enclosing the
SL and StV is illustrated by the midmodiolar 2D-sectional label
map in Figure 5(a).
A 3D model of the full cochlea combining the segmented
intracochlear structures is shown in Figure 5(b). The model
illustrates the clear separation between the cochlear ducts

C 2016 The Authors

Journal of Microscopy 
C 2016 Royal Microscopical Society, 00, 1–9
6 M. ELFARNAWANY ET AL.

Fig. 3. Improved soft tissue contrast in SR-PCI images over conventional micro-CT images. The BM has better contrast within the marked region-of-
interest (solid-line box) in the SR-PCI image (A) than the conventional micro-CT image (B). This is illustrated by the larger difference in signal intensity
(solid curve for SR-PCI and dashed curve for conventional micro-CT) in the intensity profile curves (C) corresponding to the vertical dotted lines through
the region-of-interest.

CNR Comparison visualization and contrast of thin soft tissues such as the BM,
*** **** **** SL and StV (Fig. 1). These soft tissues are more difficult to be
25
depicted by conventional absorption-based micro-CT imaging
20 because of the presence of fine, low-absorption connective tis-
sue cells and extracellular matrix (Merchant & Nadol, 2010).
Mean CNR

15 SR-PCI
micro-CT
10
5 decreasing thickness and stiffness from cochlear base to apex
0 results in the need to interpolate or extend the BM from the
Basal Middle Apical
Cochlear Turn
Fig. 4. Contrast-to-noise (CNR) ratio comparison between SR-PCI images
and conventional micro-CT images. There is statistically significant higher
CNR of the BM in SR-PCI images over conventional micro-CT at the basal,
middle and apical cochlear turns.
(SV, ST) and clear visualization of the SL and StV all the way
to the apical turn. Figure 5(c) shows the BM as one continuous
sheath from the base of the cochlea to its apex.
Discussion
This paper is the first work reporting the visualization of un-
stained human cochlear microstructures using SR in-line PCI.
In addition, we presented a 3D model obtained by semiauto-
mated segmentation of the complete turns of the scalae (ST
and SV) and of the BM starting from the cochlear base and
ending at the apex. The imaging and modelling methods pre-
sented here can be used to develop an atlas for segmentation
of clinical CT images. Results suggest that SR-PCI provides
superior visualization of intracochlear microstructures over
conventional absorption micro-CT.
The combination of high energy X-rays from SR and the edge
enhancement achieved by in-line PCI resulted in improved
In addition, the higher probability to miss the BM in micro-
CT images of higher cross sections of the cochlea (towards
the apex) is expected as the BM is characterized by having
(Liu et al., 2015). The inability to visualize these soft tissues
clearly visualized osseous spiral lamina on one end of the BM
to the outer bony wall of the cochlea at the other end (Whiting
et al., 2008; Noble et al., 2011). By contrast, the BM is clearly
visible from the cochlear base to apex in our SR-PCI images.
In one sample, RM was visible and the scala media could be
delineated from the SV. In the other two samples, the upper
border of the scala media could not be identified because RM
was not visible. There are two possible reasons why the RM was
not visible in these cases. First, the temporal bone around the
cochlea may have been thicker than the bone in the RM-visible
sample, making the penetration by X-rays through the sam-
ples more challenging. In these cases, increasing the energy
of the SR would improve the visualization of RM. Second, it is
also possible that the RM was damaged or broken in these two
samples during preparation. The RM may have experienced
shrinkage or was broken as the temporal bones were drilled
to fit the micro-CT sample holder. Shrinkage and breakage
have been reported in the literature as reasons for the inabil-
ity to reproducibly visualize the RM in high-resolution images
(Shibata et al., 2009).
The edge enhancement provided by the phase contrast
method and the depth of penetration provided by SR makes SR-
PCI images sharper than corresponding conventional micro-
CT images that have the same spatial resolution. Indeed, the
two imaging setups used in this work had similar spatial res-
olutions of approximately 15 µm. The edge-sharpening capa-
bility of SR-PCI allows it to depict boundaries of soft tissues

C 2016 The Authors

Journal of Microscopy 
C 2016 Royal Microscopical Society, 00, 1–9
 MICRO-CT Vs. SYNCHROTRON RADIATION PHASE CONTRAST IMAGING
Fig. 5. Sample segmentation and 3D model of intracochlear structures using SR-PCI images and the FastGrowCut algorithm. A sample label map of
a complete segmentation of the SV (yellow), ST (blue), SL (orange), StV and BM (green) (A) and the corresponding reconstructed 3D model (B) with a
separate view of the segmented BM (C).
such as nerve fibres, SL connective tissue cells and the shape
and points of contact of fine ligaments such as the annular
ligament connecting the stapes footplate to the oval window
(Fig. 2). These improvements in visualization and sharpness
are directly translated to the increased signal intensity con-
trast and the 2-fold increased CNRs illustrated in Figures 3
and 4.
The improved contrast and microstructure resolution of SR-
PCI images allows for very effective semiautomatic region-
growing segmentation. The clear and accurate segmentations
resulted in realistic and detailed 3D models of the intracochlear
structures such as the scalae and membranes (Fig. 5).
In the SR-based X-ray CT imaging literature of the
human cochlea, work presenting images with comparable in-
tracochlear microstructure visualization and soft tissue con-
trast to our images was only achieved in stained human
cochleae. Lareida et al. performed absorption-based contrast
imaging using two types of SR beams (Beamline BW2 source
with 4.2 µm pixel size that produced 10.8 keV photon en-
ergy and Beamline Tomcat source with 1.75 µm pixel size
and 12 keV energy) (Lareida et al., 2009). However, their
samples were stained using osmium tetroxide which provided
sufficient contrast to identify membranes, e.g. BM, RM, StV,
TM as well as individual ganglion cells. Another study was
performed by Muller et al. on stained human and murine
cochleae in which they utilized differential absorption con-
trast imaging and the monochromatic property of SR imaging
to acquire high-resolution images of intracochlear structures
(Müller et al., 2006). They acquired two sets of images at
energy levels above and below the absorption threshold of os-
mium tetroxide and then subtracted these images after coreg-
istration. In their stain-enhanced images, they were able to
observe the RM, BM, OC as well as the TM. In addition to os-
mium tetroxide being an expensive and very toxic chemical,
it has been reported that staining specimens causes distortion
and shrinkage of fine membranes due to dehydration (Brown
et al., 2002).
Apart from absorption-based contrast, SR-PCI techniques
were only used to image cochlear microstructures in animals.
C 2016 The Authors
Journal of Microscopy 
C 2016 Royal Microscopical Society, 00, 1–9
7
Rau et al. used a 25 keV SR source to acquire images of excised,
fixed and partially decalcified cochleae by applying SR-PCI
(Rau et al., 2007; Rau et al., 2012). They were able to visualize
ST, SV, RM, SG, BM, TM, OC and CN. The partial decalci-
fication required placing specimens in 10% ethylene glycol
tetraacetic acid for 1 day. In a similar study, Shu et al. were
able to visualize ST, SV and BM of a guinea pig cochlea using
diffraction-enhanced PCI with hard X-rays (Shu et al., 2007).
Diffraction-enhanced PCI requires the use of an additional
analyser crystal between the sample and detector. This adds
to the complexity and the stability requirements of the imag-
ing setup when compared to the simple in-line propagation
PCI used in this study. Richter et al. used SR-PCI with grating
interferometers to visualize the BM, OC and Rosenthal’s canal
as well as the SG nerves of mouse cochleae (Richter et al.,
2009). Their images were acquired from 20 µm thickness
slices which required an intensive sample sectioning protocol
and limited the visualization of intracochlear microstructures
to two-dimensional reconstructed images. A summary of the
literature on SR imaging of mammalian cochleae is given in
Table 1.
The SR-PCI images combined with the segmentation ap-
proach presented in this study could be used as a basis for
developing an atlas-based segmentation method for postop-
erative evaluation of CI surgery. An atlas constructed using
these images will have the unique advantage of accurately
depicting the BM from cochlear base to apex. Previous micro-
CT-based atlases required prediction or interpolation of the
nonvisible portions of the BM during manual segmentation of
the SV and ST (Noble et al., 2011; Braun et al., 2012; Kjer,
2016) or using histological slices to exactly map the location
of the BM (Bellos et al., 2014). In addition, using SR-PCI for
image acquisition does not require any special sample prepa-
ration techniques such as extensive removal of surrounding
bone, decalcification, removal of cochlear fluids or staining.
Another advantage of SR-PCI over conventional micro-CT is
low acquisition time. In our study, each micro-CT image re-
quired 7 h, whereas the acquisition time for SR-PCI images
of the whole cochlea was 30 min. In other studies, it was
8 M. ELFARNAWANY ET AL.

Table 1. Summary of literature on SR imaging of mammalian cochleae.

Authors Specimen/preparation Imaging technique Identified structures Spatial resolution

Lareida et al. Human: stained, fixed and Absorption contrast, Cellular structures, fibre BW2: 4.3 µm Tomcat:
dehydrated Beamline BW2 and nerves, STa , SVb , SMc 1.75 µm
Beamline Tomcat SR RMd , TMe , OCf , BMg ,
StVh
Muller et al. Human: Stained, fixed and Differential Absorption ST, SV, SM, RM, BM, OC, 2.8 µm
dehydrated contrast TM
measurement, SR
Rau et al. Mouse: Fixed and decalcified SR-PCI (In-line ST, SV, SM, RM, TM, OC 1.22 µm
propagation)
Shu et al. Pig: Fixed SR-PCI (Diffraction ST, SV, BM 10.9 µm
Enhanced Imaging)
Richter et al. Mouse: Fixed and sectioned SR-PCI (Grating ST, SV, BM 0.45 µm
interferometer)
a Scala tympani,
b scala vestibuli,
c scala media,
d Reissner’s membrane,
e tectorial membrane,
f organ of Corti,
g basilar membrane, and
h stria vascularis.

reported that long integration times were chosen to minimize References

noise effects, which yielded scan times of up to 36 h per speci-
men (Braun et al., 2012).
Conclusion
SR-PCI can be used as an improved tool to visualize and
model intracochlear microstructures and membranes with-
out the need for staining or decalcification. It provides re-
markably high depiction of structural details and soft tis-
sue contrast as well as higher CNRs than conventional
absorption-based micro-CT. Semiautomatic segmentations of
intracochlear structures (SV, ST and BM) are possible using
these images and yield accurate reconstructions of 3D mod-
els of the intracochlear anatomy. These segmentations and
models represent a very promising basis for developing robust
atlas-based segmentation method for postoperative evaluation
of CI surgery.
Acknowledgements
Financial support for Mai Elfarnawany, Ph.D. was pro-
vided by MITACS Accelerate postdoctoral internship num-
ber: IT04392, Canada. Part of the research described in this
paper was performed at the BMIT facility at the Canadian
Light Source, which is funded by the Canada Foundation for
Innovation, the Natural Sciences and Engineering Research
Council of Canada, the National Research Council Canada,
the Canadian Institutes of Health Research, the Government
of Saskatchewan, Western Economic Diversification Canada
and the University of Saskatchewan.
Adunka, O., Kiefer, J., Unkelbach, M.H., Radeloff, A. & Gstoettner, W.
(2005) Evaluating cochlear implant trauma to the scala vestibuli. Clin.
Otolaryngol. 30(2), 121–127.
Adunka, O.F. & Buchman, C.A. (2007) Scala tympani cochleostomy I:
results of a survey. Laryngoscope 117(12), 2187–2194.
Bartels, M., Hernandez, V.H., Krenkel, M., Moser, T. & Salditt, T. (2013)
Phase contrast tomography of the mouse cochlea at microfocus x-ray
sources. Appl. Phys. Lett. 103(8), 1–5.
Bellos, C., Rigas, G., Spiridon, I.F., et al. (2014) Reconstruction of cochlea
based on micro-CT and histological images of the human inner ear.
BioMed Res. Int. 2014, 485783.
Braun, K., Böhnke, F. & Stark, T. (2012) Three-dimensional representa-
tion of the human cochlea using micro-computed tomography data:
presenting an anatomical model for further numerical calculations.
Acta Otolaryngol. 132(6), 603–613.
Brown, M.A., Reed, R.B. & Henry, R.W. (2002) Effects of dehydration
mediums and temperature on total dehydration time and tissue shrink-
age. J. Int. Soc. Plastination 33, 28–33.
Brunschwig, A.S. & Salt, A.N. (1997) Fixation-induced shrinkage of Reiss-
ner’s membrane and its potential influence on the assessment of en-
dolymph volume. Hear Res. 114(1–2), 62–68.
Bushberg, J.T., Seibert, J.A., Leidholdt, E.M., Boone, J.M. & Goldschmidt,
E.J. (2003) Image quality. The Essential Physics of Medical Imaging, p.
91. Lippincott Williams & Wilkins, Philadelphia, PA.
Buytaert, J.A.N., Johnson, S.B., Dierick, M., Salih, W.H.M. & Santi, P.A.
(2013) MicroCT versus sTSLIM 3D imaging of the mouse cochlea. J.
Histochem. Cyctochem. 61(5), 382–395.
Edge, R.M., Evans, B.N., Pearce, M., Richter, C.P., Hu, X. & Dallos, P.
(1998) Morphology of the unfixed cochlea. Hear Res. 124(1–2), 1–16.
Fedorov, A., Beichel, R., Kalpathy-Cramer, J., et al. (2012) 3D Slicer as
an image computing platform for the Quantitative Imaging Network.
Magn. Reson. Imaging 30(9), 1323–1341.

C 2016 The Authors

Journal of Microscopy 
C 2016 Royal Microscopical Society, 00, 1–9
 MICRO-CT Vs. SYNCHROTRON RADIATION PHASE CONTRAST IMAGING
Feldkamp, L.A., Davis, L.C. & Kress, J.W. (1984) Practical cone-beam
algorithm. J. Opt. Soc. Am. A 1(6), 612–619.
Hamzavi, J. & Arnoldner, C. (2006) Effect of deep insertion of the cochlear
implant electrode array on pitch estimation and speech perception. Acta
Otolaryngol. 126(11), 1182–1187.
Kjer, H.M. (2016) Modelling of the Human Inner Ear Anatomy and Variability
for Cochlear Implant Applications. Ph.D. Thesis, Technical University of
Denmark (DTU).
Lareida, A., Beckmann, F., Schrott-Fischer, A., Gluekert, R., Freysinger,
W. & Müller, B. (2009) High-resolution X-ray tomography of the
human inner ear: synchrotron radiation-based study of nerve fi-
bre bundles, membranes and ganglion cells. J. Microsc. 234(1), 95–
102.
Liu, W., Atturo, F., Aldaya, R., et al. (2015) Macromolecular organization
and fine structure of the human basilar membrane -RELEVANCE for
cochlear implantation. Cell Tissue Res. 360, 245–262.
Merchant, S.N. & Nadol, J.B. (2010) Schuknecht’s Pathology of the Ear.
People’s Medical Publishing House, Stamford, CT.
Metscher, B.D. (2009) MicroCT for comparative morphology: simple stain-
ing methods allow high-contrast 3D imaging of diverse non-mineralized
animal tissues. BMC Physiol. 9, 11.
Müller, B., Lareida, A., Beckmann, F., Diakov, G.M., Kral, F., Schwarm,
F. & Stoffner, R. (2006) Anatomy of the murine and human cochlea
visualized at the cellular level by synchrotron-radiation-based micro-
computed tomography. Proc. SPIE 6318, 1–9.
Noble, J.H., Labadie, R.F., Majdani, O. & Dawant, B.M. (2011) Auto-
matic segmentation of intra-cochlear anatomy in conventional CT.
IEEE Trans. Biomed. Eng. 58(9), 2625–2632.
Noble, J.H., Labadie, R.F., Gifford, R.H. & Dawant, B.M. (2013) Image-
guidance enables new methods for customizing cochlear implant stim-
ulation strategies. IEEE Trans. Neural Syst. Rehabil. Eng. 21(5), 820–
829.
Postnov, A, Zarowski, A, De Clerck, N., Vanpoucke, F., Offeciers, F.E., Van
Dyck, D. & Peeters, S. (2006) High resolution micro-CT scanning as an
innovative tool for evaluation of the surgical positioning of cochlear
implant electrodes. Acta Otolaryngol. 126(5), 467–474.
Rau, C., Crecea, V., Richter, C.-P., et al. (2007) Imaging of micro-and
nano-structures with hard X-rays. Micro Nano Lett. 2(1), 1–5.
Rau, C., Hwang, M., Lee, W.K. & Richter, C.P. (2012) Quantitative X-ray
tomography of the mouse cochlea. PLoS One 7(4), e33568.
Richter, C.P., Shintani-Smith, S., Fishman, A., David, C., Robinson, I. &
Rau, C. (2009) Imaging of cochlear tissue with a grating interferometer
and hard X-rays. Microsc. Res. Technol. 72(12), 902–907.
C 2016 The Authors
Journal of Microscopy 
C 2016 Royal Microscopical Society, 00, 1–9
9
Schindelin, J., Arganda-Carreras, I., Frise, E., et al. (2012) Fiji: an open-
source platform for biological-image analysis. Nat. Methods 9(7), 676–
682.
Shibata, T., Matsumoto, S., Agishi, T. & Nagano, T. (2009) Visualization
of Reissner membrane and the spiral ganglion in human fetal cochlea
by micro-computed tomography. Am. J. Otolaryngol. Head Neck Med.
Sur. 30(2), 112–120.
Shu, H., Zhu, P., Chen, B., et al. (2007) Investigation of biomedical in-
ner microstructures with hard X-ray phase-contrast imaging. Nucl.
Instrum. Methods Phys. Res. A. 580, 610–613.
Skinner, M.W., Holden, T.A., Whiting, B.R., et al. (2007) In vivo esti-
mates of the position of advanced bionics electrode arrays in the human
cochlea. Ann. Otol. Rhinol. Laryngol. 116(4), 1–24.
Skinner, M.W., Ketten, D.R., Vannier, M.W., Gates, G.A., Yoffie, R.L.
& Kalender, W.A. (1994) Determination of the position of nucleus
cochlear implant electrodes in the inner ear. Am. J. Otol. 15(5), 644–
651.
Teymouri, J., Hullar, T.E., Holden, T.A. & Chole, R.A. (2011) Verification
of computed tomographic estimates of cochlear implant array position:
a micro-CT and histological analysis. Otol. Neurotol. 32(6), 980–986.
Verbist, B.M., Ferrarini, L., Briaire, J.J., et al. (2009) Anatomic considera-
tions of cochlear morphology and its implications for insertion trauma
in cochlear implant surgery. Otol. Neurotol. 30(4), 471–477.
Vezhnevets, V. & Konouchine, V. (2005) “GrowCut”-interactive multi-
label N-D image segmentation by cellular automata. In Proceedings of
Graphicon. pp. 150–156.
Whiting, B.R., Holden, T.A., Brunsden, B.S., Finley, C.C. & Skinner, M.W.
(2008) Use of computed tomography scans for cochlear implants. J.
Digit. Imaging 21(3), 323–328.
Wysokinski, T.W., Chapman, D., Adams, G., Renier, M., Suortti, P. &
Thomlinson, W. (2015) Beamlines of the biomedical imaging and ther-
apy facility at the Canadian light source – part 3. Nucl. Instrum. Methods
Phys. Res. A 775, 1–4.
Xu, J., Stevenson, A.W., Gao, D., et al. (2001) The role of radiographic
phase-contrast imaging in the development of intracochlear electrode
arrays. Otol. Neurotol. 22(4), 862–868.
Zhu, L., Kolesov, I., Gao, Y., Kikinis, R. & Tannenbaum, A. (2014) An effec-
tive interactive medical image segmentation method using fast grow-
cut. In Proceedings of Medical Image Computing and Computer Assisted
Intervention (MICCAI), Interactive Medical Image Computing Workshop.
Zrunek, M., Lischka, M., Hochmair-Desoyer, I. & Burian, K. (1980) Dimen-
sions of the scala tympani in relation to the diameters of multichannel
electrodes. Arch. Otorhinolaryngol. 229, 159–165.