Application of Micro-Computed Tomography With Iodine Staining to Cardiac Imaging, Segmentation, and Computational Model Development

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

> TMI-2012-0082.R1 <

1

Abstract— Micro-computed tomography (micro-CT) has been

widely used to generate high-resolution 3D tissue images from

small animals non-destructively, especially for mineralized

skeletal tissues. However, its application to the analysis of soft

cardiovascular tissues has been limited by poor inter-tissue

contrast. Recent ex vivo studies have shown that contrast between

muscular and connective tissue in micro-CT images can be

enhanced by staining with iodine. In the present study, we apply

this novel technique for imaging of cardiovascular structures in

canine hearts. We optimize the method to obtain high resolution

X-ray micro-CT images of the canine atria and its distinctive

regions - including the Bachmann's bundle, atrioventricular node,

pulmonary arteries and veins - with clear inter-tissue contrast.

The imaging results are used to reconstruct and segment the

detailed 3D geometry of the atria. Structure tensor analysis shows

that the arrangement of atrial fibers can also be characterized

using the enhanced micro-CT images, as iodine preferentially

accumulates within the muscular fibers rather than in connective

Manuscript received July 13, 2012. This work was supported by a project

grant from the British Heart Foundation (PG/10/69/28524) and partially by

grants from the Engineering and Physical Sciences Research Council

(EP/F007906 and EP/I029826/1), all UK. T. Nikolaidou was supported by a

fellowship to from the NIHR Manchester Biomedical Research Centre, UK.

O. V. Aslanidi and T. Nikolaidou equally contributed to this work.

O. V. Aslanidi is with the Department of Biomedical Engineering,

Division of Imaging Sciences & Biomedical Engineering, King's College

London, London SE1 7EH, UK (Corresponding author - phone: +44-20-718-

87188; fax: +44-20-718- 85442; e-mail: [email protected]).

T. Nikolaidou and M. R. Boyett are with the Faculty of Medical & Human

Sciences, University of Manchester, Manchester M13 9NT, UK (e-mails:

[email protected] and [email protected]).

J. Zhao and B. H. Smaill are with Auckland Bioengineering Institute,

University of Auckland, Auckland 1010, New Zealand (e-mails:

[email protected] and [email protected]).

S. H. Gilbert is with L'Institut de Rythmologie et modelisation Cardiaque,

Centre de Recherche Cardio-Thoracique, Universite Bordeaux Segalen,

33076 Bordeaux, France (email: [email protected]).

A.V. Holden is with the Institute of Membrane & Systems Biology,

University of Leeds, Leeds LS2 9JT, UK (emails: [email protected]).

T. Lowe and P. J. Withers are with the Henry Moseley X-ray Imaging

Facility, University of Manchester, Manchester M13 9NT, UK (e-mails:

[email protected] and [email protected]).

R. S. Stephenson and J. C. Jarvis are with the Institute of Ageing &

Chronic Disease, University of Liverpool, Liverpool L69 3GA, UK (e-mails:

[email protected] and [email protected]).

J. C. Hancox is with the School of Physiology & Pharmacology, University

of Bristol, Bristol BS8 1TD, UK (email: [email protected]).

H. Zhang is with the School of Physics & Astronomy, University of

Manchester, M13 9PL, UK (e-mail: [email protected]).

Copyright (c) 2010 IEEE. Personal use of this material is permitted.

However, permission to use this material for any other purposes must be

obtained from the IEEE by sending a request to [email protected].

tissues. This novel technique can be particularly useful in non-

destructive imaging of 3D cardiac architectures from large

animals and humans, due to the combination of relatively high

speed (~1 hour per scan of the large canine heart) and high voxel

resolution (36 µm) provided. In summary, contrast micro-CT

facilitates fast and non-destructive imaging and segmenting of

detailed 3D cardiovascular geometries, as well as measuring fiber

orientation, which are crucial in constructing biophysically

detailed computational cardiac models.

Index Terms—X-ray imaging and computed tomography,

Heart, Animal models and imaging, Tissue modelling.

I. INTRODUCTION

TRUCTURAL information, such as cardiovascular tissue

geometry and fiber architecture, is important for

understanding the associated function, both in health and

disease [1, 2]. Thus, the distinctive conduction pathways

formed by cardiac bundles determine the sequence of electrical

activation of the healthy heart: the normal activation starts in

the sinus node, spreads to the right atrium (RA) and through

the muscular Bachmann's bundle (BB) into the left atrium

(LA), and then through the atrioventricular node (AVN) into

the ventricles. Mapping 3D architecture of the myocardium is

also important in relation to cardiac arrhythmias, and can

guide clinical ablation treatments. For example, myocardial

sleeves of the pulmonary veins (PVs) in the LA are recognized

as primary sources of ectopic electrical activity during the

most common cardiac arrhythmia, atrial fibrillation (AF), and

ablation of the PVs is widely used to terminate AF [3, 4].

Connections between the atria via the BB allow fast

synchronized atrial activation in healthy hearts [2], but can

also be involved in atrial arrhythmogenesis [5]. Therefore,

functional studies of atrial conduction and AF require detailed

structural reconstruction of both atria, the BB and PV sleeves,

as well as the AVN which protects the ventricles from fast

atrial rates during AF and other arrhythmias [6].

Understanding of the electrophysiological mechanisms

underlying cardiac arrhythmias often emerges from

biophysically detailed computational models [7, 8] that use

structural data as the source of 3D computational domains

(geometry) and related sub-domains (tissue types). However,

structural complexity can make it difficult to quantify fine 3D

Application of Micro-Computed Tomography with

Iodine Staining to Cardiac Imaging, Segmentation

and Computational Model Development

Oleg V Aslanidi, Theodora Nikolaidou, Jichao Zhao, Bruce H Smaill, Stephen H Gilbert,

Arun V Holden, Tristan Lowe, Philip J Withers, Robert S Stephenson, Jonathan C Jarvis,

Jules C Hancox, Mark R Boyett and Henggui Zhang

S

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

> TMI-2012-0082.R1 <

2

details, such as distributions of tissue types and fiber

orientation. Thus, even the most detailed 3D models of atrial

activation have not included full descriptions of tissue

architecture throughout the atrial chambers, although some

have incorporated prescribed bundle anisotropy to account for

the role of specialized conduction pathways [9, 10].

Although both experimental functional studies and

computational modelling require accurate imaging of 3D

structures, few widely applicable methods exist for non-

destructive whole-volume imaging of soft tissues. The most

established method for imaging soft tissues has been

histological sectioning [7, 11, 12]. Magnetic resonance

imaging (MRI) methods, such as diffusion tensor MRI, gained

in prominence for reconstructing the tissue geometry and fiber

architecture non-destructively [13-16]. However, such

techniques are relatively slow. An alternative non-destructive

method of X-ray micro-computed tomography (micro-CT) has

shorter acquisition times and is widely used for imaging

diverse mineralized tissues [17, 18], but has been disregarded

in soft tissue imaging due to the poor inter-tissue contrast.

X-ray contrast enhancement agents are used routinely in

clinical radiography, but only recently have been shown to

allow quantitative characterization of soft tissues ex vivo.

Thus, Metscher [19] has used micro-CT to visualize fine soft-

tissue detail in embryos stained with iodine: the radio-opaque

staining varied the tissue density and thereby resulted in the

differential attenuation of X-rays. Micro-CT with iodine

staining has also been used for imaging of cardiac geometries

at various stages of embryogenesis in small animals [20].

Iodine staining of smooth muscular tissues has recently

produced clear inter-tissue contrast in micro-CT images, as

iodine preferentially accumulated in muscular fibers rather

than in connective tissue [21]. Finally, Stephenson et al. [22]

for the first time applied the contrast micro-CT to reconstruct

cardiac structures, primarily the cardiac conduction system in

small animal hearts.

We apply the contrast micro-CT for high resolution imaging

of cardiovascular architectures in a large canine heart -

primarily, to reconstruct structures involved in the electrical

activation of the atria. Such a reconstruction can be used to

create detailed 3D computational models of the atria and

explore electro-anatomical factors responsible for the

development and maintenance of AF.

II. METHODS

A. Tissue preparation

The heart was removed after euthanasia of a healthy adult

female boxer dog, ~8 years old, and body weight 36 kg. The

dog was euthanized with pentobarbital sodium and its body

was donated to Glasgow Veterinary School in accordance with

UK Veterinary Surgeons Act (1966). The body was

immediately chilled to 4°C, and the heart was removed within

24 hours. Heart removal was with attached lung, to preserve

PV anatomy. The heart was washed in saline and immersion

fixed in 10% neutral buffered formaldehyde (NBF, Sigma-

Aldrich). The heart was stored immersed in NBF until

imaging. To study atrial anatomy and fiber orientation, dog

atria were dissected and stained using a method described

recently [22]. Staining was optimized in 8 sequential

experiments with the same tissue sample and varying

concentrations of iodine potassium iodide solution (5-10%

I2KI) and also varying duration of sample incubation in the

solution (4-7 days). Each time the tissue sample was stained

and micro-CT imaged, after which the contrast agent was

leached out by placing the tissue in NBF for at least a week.

The leached sample was afterwards re-stained and re-scanned.

Optimal contrast of micro-CT images was achieved after 7

days of incubation with 7.5% I2KI. After staining, the tissue

was rinsed with NBF, excess solution drained and the sample

was mounted in a plastic container onto the rotatory micro-CT

scanner stage.

B. Micro-CT scanning

Samples were scanned using a Nikon Metris 225/320 KV

housed in a customized bay system at the Henry Moseley X-

ray imaging facility, University of Manchester. During the

analysis the specimen was rotated through 360 degrees and the

projections were recorded on a 2K x 2K Perkin Elmer 1621-

16-bit amorphous silicon flat-panel detector with 200 pixel

pitch. X-ray beam energy was adjusted to optimize resolution

using a Mo-target, Cu-filter (thickness 0.5 mm) combination.

The following settings were used for the analysis: scanning

time 60 minutes, voltage 150 KV, current 125 µA, gain 16. As

a result, 2001 projections per specimen were collected using a

frame rate of 2000 ms and a voxel resolution of 36 µm.

Relation between the voxel resolution and the effective tissue

resolution that can be used for computational purposes is

discussed below (see Discussion).

C. Tissue segmentation

Post processing of the raw micro-CT data included its

reconstruction using Nikon Metrolasis CT-Pro software

(Metris XT 1.6) and visualization using Avizo 6.3.1 standard

edition. Segmentation based on the iodine-enhanced inter-

tissue contrast and subsequent volume rendering was used to

reconstruct 3D atrial structures. Briefly, areas with distinctive

fiber structure, such as atrial walls and myocardial bands of the

BB and PV sleeves, had relatively high contrast in the

acquired micro-CT images due to the preferential

accumulation of iodine within the fibers. Such continuous

high-contrast areas were tracked and segmented using the

semi-automatic 'Confidence Connected' method in Avizo. In

case when the semi-automatic tracking was not effective - for

example, due to small tissue heterogeneity or sharp edges -

manual image-by-image segmentation aimed at tracking over

such heterogeneities was applied. The AVN was segmented

manually through a series of images, as it stained differentially

from the surrounding connective and myocardial tissues.

D. Structure tensor analysis

Grayscale intensity gradient information obtained from

imaging can relate the structure of objects in an image to

features of interest, e.g., the long axis of myocytes. The

structure tensor method, representing gradient information for

3D imaging problems [12, 23], was implemented. The

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

> TMI-2012-0082.R1 <

3

structure tensor contains gradient information at each voxel in

3D volume in form of a matrix. The tensor field is smoothed

by convolving it with a Gaussian kernel before an eigenvalue

decomposition solver is applied to obtain the eigenvalues and

eigenvectors of the tensor field. Local fiber alignment is

modelled as the orientation with the least signal variation,

which corresponds to the eigenvector paired with the smallest

eigenvalue. At the final stage, the fiber field is further

smoothed by averaging fiber orientations in the neighborhood

of each voxel. State of the art reconstruction methods for atrial

fiber tracking using the structure tensor have been developed

previously [12]. A simple line interpolation algorithm was

used: the 3D trajectory propagates along a line starting from a

seed point in a region of interest with a predefined sub-pixel

size and varying vector orientation, and the vector orientation

keeps itself updating by averaging eight neighboring vectors in

any new coordinate during propagation.

III. RESULTS

Fig. 1 shows the 3D tissue geometry of a canine heart

reconstructed from micro-CT images. High resolution

achieved with micro-CT allows for the high level of detail in

the 3D geometry, such as fine structures of separate blood

vessels (Fig. 1A) and pectinate muscles (PMs) in the right

atrial appendage (RAA) of the heart (Fig. 1B). Note that

micro-CT imaging of the heart at 36 µm voxel resolution took

about ~1 hour, which is considerably faster than MRI

techniques - for comparison, 300 µm reconstruction of the

same heart on 3T medical MRI scanner took ~16 hours.

Importantly, the 3D tissue geometry can be reconstructed from

micro-CT either with or without iodine staining.

Atria of the heart seen in Fig. 1 were dissected by removing

lung tissue and most of the ventricles (Fig. 2A) - however,

parts of the ventricles were kept in order to preserve the AVN.

The resultant atrial sample was stained with iodine and

scanned to produce micro-CT images with both high voxel

resolution (36 µm) and high inter-tissue contrast (Fig. 2B).

The images and the knowledge of well-known anatomic

features allowed us to (i) segment both atrial chambers and

large blood vessels, (ii) segment smaller, but distinctive atrial

structures - the BB, PVs and AVN - that play important roles

in AF, and (iii) reconstruct fiber orientation in these structures.

A. Segmentation of blood vessels

The aorta (Ao) and pulmonary arteries (PAs) are large

blood vessels that carry blood from the left and right

ventricles, respectively. Smooth muscles forming the vascular

walls (i) can be easily stained with iodine [21] and (ii) are

anatomically distinctive from myocardial tissues of the atria

[24]. This enabled semi-automatic reconstruction of the blood

vessels from the contrast micro-CT images (Fig. 3). Fig. 3A

shows the segmented volumetric 3D structures of Ao and PAs,

and Fig. 3B illustrates the identification of these structures

based on inter-tissue contrast in the images. Thus, micro-CT

with iodine staining provides relatively fast and easy means for

reconstructing geometries of blood vessels, which are in good

agreement with more established but more time demanding

PMsRAA

A B

1 cm

RV

Lung

Fig. 1. 3D geometry of the canine heart reconstructed from micro-CT. A:

Epicardial view. B: Endocardial view. Isosurfaces of the X-ray intensity in

micro-CT images are shown, with isovalues chosen such that the epicardial

and endocardial surfaces of the heart are seen. Fine tissue structures, such as

small blood vessels in the RV (A) and the network of PMs inside the RAA

(B) can be resolved from the high resolution (voxel size of 36 m) images.

RAA, right atrial appendage; RV, right ventricle; PMs, pectinate muscles.

RAALA

RV

A

Ao

PA

1 cm

RAALA

RV

B

Ao

PA

1 cm

Fig. 2. Non-segmented 3D geometry of the canine atria. A: 3D volume

rendering of the dissected atria (X-ray intensity is color-coded using a "hot

metal" palette) shows the direct 3D volume renderings - this is also. Details

of anatomical structures are not clear, for example large Ao is obscured by

other tissues. B: Respective micro-CT images (two orthogonal planes x-y

and x-z) with clear inter-tissue contrast between structures, which enables

segmentation of the 3D geometries (see below). Micro-CT image brightness

in A and B is related to the X-ray intensity: brighter colors correspond to

lower intensity/higher absorption of X-rays, which is due to higher levels of

iodine accumulation in the respective tissues. Ao, aorta; PA, pulmonary

artery; RAA, right atrial appendage; LA, left atrium, RV, right ventricle.

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

> TMI-2012-0082.R1 <

4

anatomical studies [2, 24]. Note that the PVs that carry blood

from the lungs to the LA were segmented separately as part of

the LA (see below). Although vascular walls of the PVs are

similar to those in the arteries, their long sleeves branching

from the LA are formed by myocardial tissue.

B. Segmentation of the atria

Fig. 4 illustrates the segmented 3D geometry of the canine

atria. The RA, LA, BB and PVs were clearly seen in the

micro-CT images (Fig. 4A) due to the accumulation of iodine

within their myocardial fibers. This enabled semi-automatic

segmentation of the atria into distinctive 3D tissue regions

(Fig. 4B, C). First, the interatrial BB was segmented based on

its anatomically-defined location and cable-like fibrous

structure. Second, branching PVs were traced as vascular

structures starting superiorly and then - as continuous

myocardial sleeves - joining the LA. Finally, the LA was

formally identified as the chamber adjacent to the already-

segmented PV region, and the RA was identified as the second

chamber separated from the LA by the BB (Fig. 4B).

Segmentation of the RA included the superior vena cava

(SVC) located superiorly (Fig. 4B) and a fine network-like

structure of conductive PMs in the RAA (Fig. 4C). The

segmented 3D atrial geometry (Fig. 4) was in good qualitative

agreement with anatomical studies [2, 24, 25].

Thus, the contrast micro-CT enabled (i) reconstruction of

Ao

PA

A

1 cm

Ao

PA

B

1 cm

Fig. 3. Segmentation of the 3D geometry of blood vessels. A: Segmented

geometry of Ao and PAs. Vascular regions are reconstructed and rendered as

3D volumetric digital masks. B: Identification of Ao and PAs (dotted lines)

in the respective high contrast micro-CT images (orthogonal planes x-y and

x-z). Similar to Fig. 2B, brighter colors in the images correspond to lower

intensity/ higher absorption of X-rays, which is due to higher accumulation

of iodine in the respective tissues. Ao, aorta; PA, pulmonary arteries.

C

A

PV

RA

LABB

B

1 cm

Fig. 4. Segmentation of 3D geometry of the canine atria. A: Identification of

the RA, BB, LA and PVs based on inter-tissue contrast in micro-CT images.

As before, brighter colors correspond to lower intensity/higher absorption of

X-rays, which is due to higher accumulation of iodine in the respective atrial

tissues. B and C: Superior and posterior views of the segmented 3D atria.

Segmented atrial regions are reconstructed and rendered as 3D volumetric

digital masks. 3D atrial regions are shown using same colors as lines in A.

Clear structure of PMs inside the RA is seen in C.

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

> TMI-2012-0082.R1 <

5

the 3D geometry of canine atria and (ii) their segmentation into

major atrial structures (tissue types) with a high level of detail.

High resolution and high contrast of the micro-CT images also

allowed for the reconstruction of fiber orientations in muscular

atrial bundles, such as the BB and PV sleeves (see below).

C. Fiber orientation in the BB

The BB is the largest atrial bundle that connects the atria

and during normal sinus rhythm provides the primary pathway

for the rapid interatrial spread of electrical activation. Despite

its importance in the synchronized atrial activation and a role

in AF arrhythmogenesis [5, 26], there are relatively few

anatomic studies of the BB [2, 24, 27]. Detailed fiber

orientation in the BB has been reconstructed only recently

using time and labor demanding histological methods [12].

The established cable-like structure and clear arrangement of

fibers make the BB an ideal object for reconstruction using the

novel fast method of micro-CT with iodine staining.

Such a reconstruction is shown in Fig. 5. The arrangement

of fibers along the BB is illustrated by a histological tissue

slice (Fig. 5A) - a similar pattern can be seen in micro-CT

images due to iodine staining of the BB fibers (Fig. 4A). The

BB segmented from these images (Fig. 5B) was used as a 3D

digital mask for applying the structure tensor method for fiber

tracking (Methods). The resultant cable-like arrangement of

fibers along the BB can be clearly seen in Fig. 5C, which is in

good qualitative agreement with the histological data from

sheep [12] and human anatomical studies [2, 24]. Although

direct quantitative comparisons between data from various

species are not straightforward, the contrast micro-CT

reconstruction of this prominent bundle in dog (Fig. 5C)

generally agrees with the histological [12] and anatomical [24]

observations: (i) the vast majority of fibers are strictly aligned

along the BB with the inclination angle of <100; (ii) a small

fraction of fibers have higher inclination angles of ~20-300.

This provides a validation for the micro-CT imaging of fiber

orientation in the canine atria.

D. Fiber orientation in the PVs

The myocardial sleeves of the PVs are recognized as the

primary source of ectopic electrical activity during AF [3, 4],

which may be due to complex arrangement of fibers in this

distinctive region of the LA [28, 29]. The myocardial sleeves

of the PVs were clearly seen in micro-CT images (Fig. 4A), as

iodine staining of their fibers resulted in high contrast with the

surrounding connective tissues. The segmented 3D structure of

the PV region is seen in Fig. 4B: four branching PVs are

located at the superior side of the LA.

Fig. 6 shows the novel micro-CT reconstruction of fiber

orientation in the PV region. Similar to the BB, the segmented

00

5 mm

B

C

900

RALA

PV

BB

A

1 cm

Fig. 5. Reconstruction of fiber orientation in the BB. A: High resolution

histological section (8 m) of the sheep atria [12]. Clear arrangement of

fibers along the BB can be seen. B: Segmented geometry of the BB (see also

Fig. 4) used as a 3D digital mask for reconstructing fiber orientation. C:

Reconstructed fiber orientation in the BB. Fibers are colored according to

their inclination angle ("rainbow" palette). Fibers are aligned in the direction

along the BB, which is in good agreement with histological studies.

A B

C D

00 900

5 mm

Fig. 6. Reconstruction of fiber orientation in the PVs. A and B: Segmented

geometry of the PV region (see also Fig. 4) used as a 3D digital mask for

reconstructing fibers. C and D: Reconstructed fiber orientation in the PVs.

Fibers are colored according to their inclination angle. Fibers are mostly

aligned along the PV sleeves, but their arrangement becomes more complex -

with multiple changing directions - towards the LA (dark arrow), which

agrees with other studies [12, 25, 28]. A characteristic pattern of fibers [25]

can be seen in the inter-pulmonary area, where circumferentially aligned

strands meet with predominantly longitudinal strands (white arrows).

Posterior (A, C) and anterior (B, D) views are shown.

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

> TMI-2012-0082.R1 <

6

PV region (Figs. 6A, B) was used as a digital 3D mask for

applying the structure tensor method for fiber tracking. The

resultant fiber orientation was in agreement with available

(however, limited) knowledge of this region [12, 25]. Fibers

were aligned along the myocardial sleeves of the PVs, but the

orientation became complex - with fibers arranged in multiple

directions and changes in the arrangement - as the PV sleeves

extended towards the LA (Figs. 6C, D). Such a complex

arrangement of fibers at the PV-LA junctions is difficult to

characterize [25]. However, characteristic changes in the

orientation of fiber strands can be seen in the inter-pulmonary

areas, where obliquely or circumferentially aligned strands

meet with predominantly longitudinally aligned strands [25].

Qualitatively similar patterns are observed in the micro-CT

reconstruction (see Fig. 6C, with arrows showing the strand

directions), as well as in histological [12] and electro-

anatomical [28] studies. Along with the recent histological

reconstruction by Zhao et al. [12], this study provides unique

high resolution information on the fiber orientation in the PVs.

E. Atrioventricular node

The AVN plays a key role in coordinating electrical

conduction from atria to the ventricles. Studies of the AVN

have been limited due to its small size, complex anatomy and

inaccessible location in depth of the atrioventricular septum [6,

30]. Micro-CT with iodine staining provides high resolution

and high contrast images, and allows for non-destructive in-

depth tissue studies. This novel method combined with the

knowledge of cardiac anatomy is perfectly suited for fast

tracing of the AVN and reconstruction of its 3D geometry.

The AVN was segmented by tracing its margins through

serial images. Its location near the thick fibrous skeleton that

connects the central fibrous body with the aortic, mitral and

tricuspid valves was clearly seen in the micro-CT images.

Image contrast allowed us to define 3D geometry of the AVN

in situ, without distorting its relation to the surrounding fine

structures (Fig. 7). Primarily, the AVN was identified at the

crest of the ventricular septum and traced posteriorly and

anteriorly along the septum based on the inter-tissue contrast

(Figs. 7A, B). The reconstructed 3D geometry was further

subdivided into the posterior nodal extension (PNE), the

compact node (CN) and AV bundle (AVB) (Fig. 7C). Note

that the subdivision was based on the knowledge from

previous studies of the AVN anatomy and function [30, 31].

Note also that a histological validation for the micro-CT

segmentation for the atrioventricular conduction axis in a small

rat heart has been performed recently [22] (see Discussion).

Thus, the contrast micro-CT enabled the reconstruction of the

3D structure of such a small (~13.0×4.0×0.5 mm3) and

complex anatomical object as the canine AVN.

IV. DISCUSSION

We have demonstrated a variety of applications of X-ray

micro-CT with iodine staining to imaging of cardiovascular

tissues. The high resolution (voxel size of 36 µm) and high

contrast imaging was used to (i) reconstruct 3D geometry of

the canine atria (Fig. 1), (ii) segment the atria into major

myocardial tissue types - RA, LA, BB and PV (Fig. 4), (iii)

segment large vascular structures - Ao and PAs (Fig. 3), (iv)

segment fine structure of the AVN (Fig. 7), and (v) reconstruct

fiber orientation of distinctive atrial bundles - the BB (Fig. 5)

and sleeves of the PV (Fig. 6). Such segmented 3D geometries

and fiber orientations are crucial in constructing biophysically

detailed computational models and applying them to study

mechanisms of arrhythmias. Thus, the detailed 3D architecture

of the atria reconstructed from micro-CT can be used to create

computational models for exploring atrial conduction and AF

arrhythmogenesis. Below we discuss potential applications of

our methods and results, as well as their limitations.

A. Micro-CT with iodine staining

X-ray micro-CT is a high resolution version of medical CT

scanning, which has been increasingly used in non-clinical

research over the last decade. A sample is placed in the path of

an X-ray beam to generate a projection image on an X-ray-

sensitive detector. After the sample is rotated and imaged at a

large number of angles, the sequence of projection images is

used to reconstruct the X-ray attenuation at each point within

the scanned 3D volume. Thus, micro-CT imaging represents

the sample as a 3D matrix of intensity values, equivalent to a

stack of aligned 2D digital images.

Advantages of micro-CT include 1) non-destructive

approach to tissue imaging, 2) relatively short image

acquisition times, 3) high resolution (up to ~1 µm) of the

images and 4) high contrast between tissues with differential

attenuation of X-rays. Therefore, the method has been widely

PNE CN

AVB

LA

RV

Ao

RA

Ao

AS

VS

AVN

AVN

A B

C D

5 mm

5 mm

Fig. 7. Reconstruction of the AVN from the contrast micro-CT images. A:

Identification of the AVN margins (dotted line) in the contrast-enhanced

images. B: Segmented 3D structure of the AVN. As in Figs. 3-6, the

segmented region is reconstructed and rendered as a 3D volumetric digital

mask. C: Anatomy of the AVN. The PNE is identified at the base of the

tricuspid valve and continues anteriorly as the CN and AVB. D:

Relationship of the segmented AVN to surrounding tissues seen in micro-CT

images. Ao, aorta; AS and VS, atrial and ventricular septum; PNE, posterior

nodal extension; CN, compact node; AVB, atrioventricular bundle; RV,

right ventricle; RA and LA, right and left atria.

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

> TMI-2012-0082.R1 <

7

used for high resolution imaging of diverse mineralized tissues

ex vivo [17, 18]. Limitations of micro-CT arise from the facts

that 1) high energies of the X-ray beam are dangerous to living

organisms, which makes in vivo studies nearly impossible, and

2) tissues with similar X-ray absorption may have poor inter-

tissue contrast. Therefore, micro-CT has until recently been

disregarded in soft tissue imaging. However, recent studies

have shown that X-ray absorption in various soft tissues can be

greatly enhanced by using contrast agents such as iodine.

Micro-CT with iodine contrast enhancement has been used to

image details of ex vivo embryonic [19, 20] and smooth

muscle [21] tissues. Recently the contrast micro-CT has been

applied to cardiac tissues [22]. In such studies, resolutions of

~1 µm can be achieved in a small field of view [18, 32].

Mechanisms for the iodine tissue staining and contrast

enhancement can be explained by its diffusion through tissue

layers and binding to glycogen within muscle cells [21, 33].

The resultant accumulation of iodine in the muscular fibers

leads to a relative increase of X-ray absorption in the fibers

compared to connective tissues. Our results demonstrate that

the combination of micro-CT with iodine staining can be used

for the direct high resolution, high contrast imaging of

cardiovascular tissues, including details of fiber architecture.

B. Comparison of micro-CT with histology

The most established method for reconstructing soft tissues

is dissection/sectioning combined with staining of the resultant

histological sections and followed by microscopic imaging.

This method has been used for reconstructing 3D geometries

(and consequently biophysically detailed models) for various

parts of the heart. These include several models for the

ventricles [7, 34, 35], atria [12, 36], sinus node [11] and AVN

[30]. Such studies are considered the reference standard for the

cardiac model development due to high resolution (achieved

by fine tissue sectioning and microscopy) and high contrast

(achieved by staining) of histological images. However, the

method has several limitations: 1) tissue sectioning is

destructive, such that each sample can be used only once, 2)

resolution within and between histological sections is limited

by two different factors (resolutions of the microscope and the

sectioning method used, respectively), 3) sectioning can result

in damage and deformation of the sample such that meaningful

data are lost, 4) variations in staining between serial tissue

sections may give rise to artificial heterogeneities, 5)

reconstruction of the 3D geometry from a stack of 2D

histological sections/images requires their alignment, which

can also lead to artifacts; 6) as most stages of tissue sectioning,

staining, imaging and reconstruction are done manually, the

method is extremely time and labor demanding.

The contrast micro-CT technique described in this study

has several potential advantages over histological methods for

studying cardiovascular architectures. Unlike sectioning, the

method is non-destructive and iodine staining is easily

reversible [22]. This means the same sample can be repeatedly

scanned to optimize the imaging of different features, which

was used in the present study with the canine atria. The non-

destructive nature of the method is an important consideration

when dealing with rare tissues (e.g., human) or unusual

pathologies, which can help improve the understanding of

cardiovascular disease. The micro-CT method is much faster:

scanning the canine atria took ~1 hour, compared to several

months required for histological studies [11, 12, 30]. Finally, it

can produce high resolution 3D images that are intrinsically

isotropic and free from tissue deformation artifacts.

C. Comparison of micro-CT with MRI

In clinical cardiac imaging MRI is the reference standard

modality for assessing cardiac anatomy and evaluating

myocardial contractile function [37]. Cardiac CT has mainly

been used for the exclusion of coronary artery disease. The

main disadvantage of CT for clinical cardiac imaging is the

exposure of the patient to ionizing radiation. Experimental

cardiac imaging has largely been developed from clinical

approaches and therefore MRI is also the dominant imaging

modality [38]. However, exposure to ionizing radiation is not

important during ex vivo studies of cardiac microstructure, and

micro-CT may have an advantage over MRI in resolution,

speed and tissue contrast. These factors are discussed below.

Associated with the unique manner in which MRI generates

an image (from radio-frequency radiation induced in water

molecules) are physical factors which limit the use of MRI for

high resolution imaging. The main limitation is the extent of

movement of the water molecules within the tissue, known as

self-diffusion: over a typical MRI pulse of ~100 ms water

molecules can diffuse over the distance of ~8 µm, hence

limiting imaging resolution to >8 µm [39]. In comparison,

micro-CT resolution is limited primarily by scanner design,

and resolutions of ~1 µm can be achieved [18].

Another limitation of MRI is related to the signal-to-noise

ratio (SNR): increasing the resolution leads to a proportional

increase of the SNR for a given scan time. SNR can be

decreased (and hence, achievable resolution increased) by

increasing imaging time. As a result, long MRI imaging

sequences are required to obtain the highest quality images.

For example, ex vivo high resolution MRI (25 × 25 × 37 µm)

of whole fixed rat hearts has required 20 signal measurements

over 72 hours and averaging the result [16]. This may be

practical for stable fixed samples, but prevents the acquisition

of high resolution MRI images from fresh tissue. Therefore,

micro-CT may have more potential for the high resolution

imaging of non-fixed cardiac tissues.

Conventional MRI images provide poor myocardial contrast

and poor resolution of fibers. This can be enhanced by using

perfused contrast agents. Targeted contrast agents have been

developed for MRI that allow labeling of tissue features of

interest, e.g., collagen in fibrotic tissue [40]. However, contrast

MRI has not been used in studies of cardiac architectures.

Diffusion tensor MRI (DT-MRI) has been widely applied

for measuring the fiber and laminar orientation from hearts ex

vivo [13-15]. Accuracy of these measurements has also been

validated against histology [13, 41]. DT-MRI provides global

cardiac architecture maps that are useful in studies of the

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

> TMI-2012-0082.R1 <

8

myocardial organization and can be applied as computational

domains in 3D cardiac modelling [8, 11, 42]. However, the

maximum resolution of DT-MRI reported in the ex vivo

myocardium is ~100 µm, with acquisition times of ~10 hours

for volumetric scans of a small mouse heart [15]. DT-MRI is

limited to this resolution as the SNR of the imaging is

inherently low [15] and the tensor estimation is highly noise-

sensitive. As a result, DT-MRI images with low SNR result in

poorly assigned fiber orientations [43].

D. Effective atrial tissue resolution

3D atrial tissue architectures were reconstructed in this

study from the contrast micro-CT data at the voxel resolution

of 36 μm. However, the latter value is not the "true" tissue

resolution that can be incorporated into computational models

as the space step of numerical integration. There is no

straightforward procedure for measuring the "true" tissue

resolution from micro-CT imaging data. This is because the

reconstructed resolution may depend upon several factors: (i)

The focal spot size on the target sample. In our experiments

this was 3 μm for the voltage of 150 KV, however the size

generally varies from experiment to experiment depending on

the used voltage value. (ii) The sample to source and sample to

detector distances. (iii) The signal to noise ratio. A clear signal

is obtained by empirically selecting the correct voltage and

current for the sample, such that a ~30% transmission of X-

rays through the sample is obtained. The signal is afterwards

maximized, which was done in our experiments through

~50000 counts on the X-ray detector. The reported voxel size

of 36 μm was calculated during the reconstruction stage by the

Nikon Metrolasis CT-Pro software, which takes into account

factors (i)-(ii), but not (iii). As an empirical rule, for the Nikon

Metris custom bay system and the signal-to-noise ratio

optimization procedure used the "true" resolution is about (or

less than) twice the voxel size. Hence, the effective resolution

of the reconstructed atrial tissue is approximately 70 μm. This

value can be used in computational modelling of the 3D atria.

Note that the integration space step of 70 μm is more than

sufficient for computational purposes. For example, Zhao et al.

[12] histologically reconstructed the 3D sheep atrial model

with the voxel resolution of 50 μm; in simulations performed

with the model, no quantitative difference between results

obtained with the space steps of 50 and 100 μm was observed.

E. Computational models of the atria

Computational models with high degree of biophysical

detail have been developed for major parts of the heart.

Primarily, anatomically detailed 3D geometries have been

developed for the ventricles utilizing histological [7, 34, 35]

and later MRI [13, 14] techniques, and used as computational

domains in cardiac function modelling [8, 42].

Models of the atria have been based on the histologically

reconstructed Visible Female human geometry [9, 10, 36].

Reconstructions based on volumetric MRI [44] and CT [45]

have also been used to obtain generic surface geometries of the

atria. While these models included various details of atrial

anatomy, their segmentation into electrophysiologically and

anatomically distinctive tissue sub-domains was either absent

[36, 44] or based on phenomenological estimations of the sub-

domain locations [9, 45]. Moreover, even the most detailed 3D

models have not included accurate descriptions of fiber

architecture in the atria. Only few models have incorporated

prescribed local bundle anisotropy to account for the role of

specialized atrial conduction pathways, such as the BB [9, 10].

The model of the human atria by Aslanidi et al. [10]

overcame many limitations of earlier models by considering

detailed 3D geometry, as well as accurate electrophysiological

heterogeneity and local anisotropy (which was partly based on

DT-MRI data [11]). The latter two features of the model were

particularly important in simulations of AF arrhythmogenesis.

Primarily, the simulations showed that tissue heterogeneity

caused the break-down of the normal activation wavefronts at

rapid pacing rates, which initiated a pair of re-entrant scroll

waves - and tissue anisotropy resulted in a further break-down

of the scrolls into multiple meandering wavelets characteristic

of AF. This provided insights into the 3D dynamics of AF in

depth of the atria, which is beyond the current technical

capabilities of experimental or clinical set-ups.

However, even the most detailed 3D atrial model [10] has

not accounted for electrophysiological heterogeneity and

anisotropy of the PV sleeves, which are crucial in the genesis

of AF [3, 28, 29]. The contrast micro-CT method used in this

study provides (i) the novel segmentation of the PV sub-

domain and (ii) reconstruction of detailed fiber orientation in

the PV sleeves. Moreover, the method reconstructs the entire

segmented 3D atria with the effective resolution of ~70 µm,

which is a great improvement over the resolution of 330 µm

provided by the widely used Visible Female dataset. A new

family of heterogeneous electrophysiological models for the

canine RA, LA, BB and PV cells has been developed recently

[46]. Currently these models are being incorporated into a new

3D computational model that integrates the atrial cell

electrophysiology with the tissue geometry and fiber

orientations reconstructed from the contrast micro-CT data.

F. Limitations

Although time and labor intensive histological experiments

were beyond the scope of this study, a histological validation

will ultimately be required in order to evaluate the accuracy of

reconstruction of structural features (such as fiber orientations

in the BB and PVs) reconstructed using micro-CT.

Currently, we can only qualitatively compare the micro-

CT reconstruction of canine atria with histological results from

sheep atria [12]. Quantitative comparisons between various

experimental studies of the atrial fibers are extremely difficult

considering the varying atrial shape, size, wall thickness and

complex fiber pattern. Thus, Zhao et al. [47] have recently

compared fibers in human and sheep atria and illustrated

qualitative similarities of the structure of large atrial bundles in

these two species. But the same study has also suggested that

quantitative comparisons between atrial micro-architectures

from two different species were virtually impossible as the

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

> TMI-2012-0082.R1 <

9

fiber information belonged to different tissue geometries.

Hence, qualitative similarities between the contrast micro-CT

reconstruction of the PVs (Fig. 6) and the respective

histological [12], anatomical [25] and electro-anatomical data

[28] may be the only validation currently available for this

region. Note that although direct quantitative comparisons

between data from various species are not straightforward,

fiber inclination angles in the prominent BB reconstructed

from the contrast micro-CT (Fig. 5) were in good agreement

with histological [12] and anatomical [24] observations.

For the AVN segmented from the contrast micro-CT

images (Fig. 7), a direct histological validation was once again

beyond the scope of the present study. However, a histological

validation for the contrast micro-CT segmentation of the

atrioventricular conduction axis in a rat heart has been done

[22]. Direct comparisons of the micro-CT and histological

images of the same AVN tissue have shown that structures

~100 μm can be distinguished, and only individual cardiac

myocytes (<20 μm) are unclear. The canine AVN segmented

in the present study is ~13000 × 4000 × 500 μm and hence

sufficiently large to be reconstructed from micro-CT images.

Note that even high resolution histology may not always

provide clear information about atrial fiber directions. For the

sheep atrial free wall, Zhao et al. [12, 47] have histologically

reconstructed fiber orientation in several segments with the

resolution of 50 μm, and concluded that the fiber micro-

architecture in the wall is highly complex, with no clear fiber

direction in many regions. This is consistent with anatomical

observations [25]. Our preliminary reconstruction of fiber

orientation in a segment of the canine atrial free wall (not

shown) also reveals a complex fiber pattern. Efforts of

quantifying atrial micro-architectures in different species and

using various techniques (and hence, providing validated data

for computational modelling) are ongoing.

V. CONCLUSION

Understanding the spatio-temporal electrical dynamics during

normal sinus rhythm and atrial arrhythmias (such as AF)

requires full in-depth access to the atria. This is extremely

difficult to implement in an experimental or clinical set-up.

Computational models of the 3D atria, which are based on

multiple imaging modalities, can provide biophysical validated

means for dissecting and explaining electrical processes

underlying the normal and arrhythmic atrial dynamics.

Recent developments in semi-automated volumetric

histology [12, 35] and MRI/DT-MRI [13-16] have enabled

faster and higher resolution imaging of cardiovascular tissues,

but the problem of efficient reconstruction of computational

domains for biophysical modelling is still not fully resolved.

Micro-CT with iodine staining can have an advantage over

other methods in non-destructive imaging of 3D cardiac

architectures from large animals and humans, due to the

unique combination of high speed (~1 hour/scan for a large

canine heart) and high voxel resolution (36 µm) provided.

However, further validation of this novel method may be

needed. As none of the existing techniques has yet delivered

the golden standard for cardiovascular imaging, a combination

of multiple mutually-validating modalities may provide the

required solution. A combination of DT-MRI, used for non-

destructive reconstruction of fiber orientation, with subsequent

histological sectioning, to obtain fine details of the tissue

structure, has been applied to create a small-scale 3D model of

the sinus node [11]. Combining contrast micro-CT, for fast

non-destructive imaging of 3D tissue geometries and fiber

orientation, with semi-automated volumetric histology and

microscopy [12], for further validation of the fiber orientation

and tissue characteristics, may provide means for effective

high-throughput generation of large-scale computational

models of cardiac chambers and the entire heart.

ACKNOWLEDGMENT

We thank Dr Maureen Bain from the School of Veterinary

Medicine at the University of Glasgow (UK) for providing and

preparing the canine heart.

REFERENCES

[1] D. Noble, "Modeling the heart – from genes to cells to the whole organ,"

Science, vol. 295, pp. 1678-82, 2002.

[2] S.Y. Ho, and D. Sanchez-Quintana, "The importance of atrial structure

and fibers," Clinical Anatomy, vol. 22, pp. 52-63, 2009.

[3] M. Haissaguerre, P. Jais, D.C. Shah, et al., "Spontaneous initiation of

atrial fibrillation by ectopic beats originating in the pulmonary veins,"

N. Engl. J. Med., vol. 339, pp. 659-66, 1998.

[4] H. Calkins, K.H. Kuck, R. Cappato, et al., "2012 HRS/EHRA/ECAS

expert consensus statement on catheter and surgical ablation of atrial

fibrillation: recommendations for patient selection, procedural

techniques, patient management and follow-up, definitions, endpoints,

and research trial design," Heart Rhythm, vol. 9, pp. 632-96, 2012.

[5] F.X. Roithinger, J. Cheng, A. Sippens-Groenewegen, R.J. Lee, L.A.

Saxon, M.M. Scheinman, and M.D. Lesh, "Use of electroanatomic

mapping to delineate transseptal atrial conduction in humans,"

Circulation, vol. 100, pp. 1791-97, 1999.

[6] T. Kurian, C. Ambrosi, W. Hucker, V.V. Fedorov, and I.R. Efimov,

"Anatomy and electrophysiology of the human AV node," Pacing &

Clin. Electrophysiol., vol. 33, pp. 754-62, 2010.

[7] P.J. Hunter, A.J. Pullan, and B.H. Smaill, "Modeling total heart

function," Ann. Rev. Biomed. Eng., vol. 5, pp. 147-77, 2003.

[8] N.P. Smith, P.J. Hunter, and D.J. Paterson, "The Cardiac Physiome: at

the heart of coupling models to measurement," Exp. Physiol., vol. 94,

pp. 469-71, 2009.

[9] G. Seemann, C. Hoper, F.B. Sachse, O. Dossel, A.V. Holden, H. Zhang,

"Heterogeneous 3-dimensional anatomical and electrophysiological

model of human atria," Phil. Trans. A, vol. 364, pp. 1465-81, 2006.

[10] O.V. Aslanidi, M.A. Colman, J. Stott, H. Dobrzynski, M.R. Boyett,

A.V. Holden, and H. Zhang, "3D virtual human atria: a computational

platform for studying clinical atrial fibrillation," Prog. Biophys. Mol.

Biol., vol. 107, pp. 156-68, 2011.

[11] N. Chandler, O.V. Aslanidi, D. Buckley, et al., "Computer three-

dimensional anatomical reconstruction of the human sinus node and a

novel paranodal area," Anat. Rec. (Hoboken), vol. 294, pp. 970-79,

2011.

[12] J. Zhao, T.D. Butters, H. Zhang, A.J. Pullan, I.J. LeGrice, G.B. Sands,

and B.H. Smaill, "An image-based model of atrial muscular

architecture: Structural anisotropy and electrical activation," Circ.

Arrhythmia Electrophysiol., vol. 5, pp. 361-70, 2012.

[13] E.W. Hsu, A.L. Muzikant, S.A. Matulevicius, R.C. Penland, and C.S.

Henriquez, "Magnetic resonance myocardial fiber-orientation mapping

with direct histological correlation," Am. J. Physiol., vol. 274, pp.

H1627-34, 1998.

[14] D.F. Scollan, A. Holmes, J. Zhang, and R.L. Winslow, "Reconstruction

of cardiac ventricular geometry and fiber orientation using magnetic

resonance imaging," Ann. Biomed. Eng., vol. 28, pp. 934-44, 2000.

Copyright (c) 2011 IEEE. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing [email protected].

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

> TMI-2012-0082.R1 <

10

[15] Y. Jiang, K. Pandya, O. Smithies, and E.W. Hsu, "Three-dimensional

diffusion tensor microscopy of fixed mouse hearts," Magn. Reson.

Med., vol. 52, pp. 453-60, 2004.

[16] S.H. Gilbert, D. Benoist, A.P. Benson, et al., "Visualization and

quantification of whole rat heart laminar structure using high-spatial

resolution contrast-enhanced MRI," Am. J. Physiol., vol. 302, pp. H287-

98, 2012.

[17] F. Neues, and M. Epple, "X-ray microcomputed tomography for the

study of biomineralized endo- and exoskeletons of animals," Chem.

Rev., vol. 108, pp. 4734-41, 2008.

[18] M.W. Westneat, J.J. Socha, and W.K. Lee, "Advances in biological

structure, function, and physiology using synchrotron X-ray imaging,"

Annu. Rev. Physiol., vol. 70, pp. 119-42, 2008.

[19] B.D. Metscher, "MicroCT for comparative morphology: simple staining

methods allow high-contrast 3D imaging of diverse non-mineralized

animal tissues," BMC Physiol., vol. 9, pp. 11-25, 2009.

[20] K. Degenhardt, A.C. Wright, D. Horng, A. Padmanabhan, and J.A.

Epstein, "Rapid 3D phenotyping of cardiovascular development in

mouse embryos by micro-CT with iodine staining," Circ. Cardiovasc.

Imaging, vol. 3, pp. 314-22, 2010.

[21] N.S. Jeffery, R.S. Stephenson, J.A. Gallagher, J.C. Jarvis, and P.G. Cox.

"Micro-computed tomography with iodine staining resolves the

arrangement of muscle fibers," J. Biomech., vol. 44, pp. 189-92, 2011.

[22] R.S. Stephenson, M.R. Boyett, G. Hart, et al., "Contrast enhanced

micro-computed tomography resolves the 3-dimensional morphology of

the cardiac conduction system in mammalian hearts," PLoS One, vol. 7,

p. e35299, 2012.

[23] M. Krause, J.M. Hausherr, B. Burgeth, C. Herrmann, and W. Krenkel,

"Determination of the fiber orientation in composites using the structure

tensor and local X-ray transform," J. Mater. Sci., vol. 45, pp. 888-96,

2010.

[24] K. Wang, S.Y. Ho, D.G. Gibson, and R.H. Anderson, "Architecture of

atrial musculature in humans," Br. Heart. J., vol. 73, pp. 559-65, 1995.

[25] S.Y. Ho, J.A. Cabrera, and D. Sanchez-Quintana, "Left atrial anatomy

revisited," Circ. Arrhythm. Electrophysiol., vol. 5, pp. 220-28, 2012.

[26] D. O'Donnell, J.P. Bourke, and S.S. Furniss, "Interatrial transseptal

electrical conduction: comparison of patients with atrial fibrillation and

normal controls," J. Cardiovasc. Electrophysiol., vol. 13, pp. 1111-17,

2002.

[27] R. Lemery, G. Guiraudon, and J.P. Veinot, "Anatomic description of

Bachmann's bundle and its relation to the atrial septum," Am. J.

Cardiol., vol. 91, pp. 1482-85, 2003.

[28] M. Hocini, S.Y. Ho, T. Kawara, et al., "Electrical conduction in canine

pulmonary veins: electrophysiological and anatomic correlation,"

Circulation, vol. 105, pp. 2442-48, 2002.

[29] P. Sanders, O. Berenfeld, M. Hocini, et al., "Spectral analysis identifies

sites of high-frequency activity maintaining atrial fibrillation in

humans," Circulation, vol. 112, pp. 789-97, 2005.

[30] J. Li, I.D. Greener, S. Inada, et al., "Computer three-dimensional

reconstruction of the atrioventricular node," Circ. Res., vol. 102, pp.

975-85, 2008.

[31] L. Sherf, T.N. James, and W.T. Woods, "Function of the atrioventricular

node considered on the basis of observed histology and fine structure,"

J. Am. Coll. Cardiol., vol. 5, pp. 770-80, 1985.

[32] B.D. Metscher, and G.B. Muller, "MicroCT for molecular imaging:

quantitative visualization of complete three-dimensional distributions of

gene products in embryonic limbs," Dev. Dyn., vol. 240, pp. 2301-18,

2011.

[33] D.N. Lecker, S. Kumari, and A. Khan, "Iodine binding capacity and

iodine binding energy of glycogen," J. Polymer. Sci. A, vol. 35, pp.

1409-12, 1997.

[34] F.J. Vetter, and A.D. McCulloch, "Three-dimensional analysis of

regional cardiac function: a model of rabbit ventricular anatomy," Prog.

Biophys. Mol. Biol., vol. 69, pp. 157-83, 1998

[35] G.B. Sands, D.A. Gerneke, D.A. Hooks, C.R. Green, B.H. Smaill, and

I.J. Legrice, "Automated imaging of extended tissue volumes using

confocal microscopy," Microsc. Res. Tech., vol. 67, pp. 227-239, 2005.

[36] D Harrild, and C. Henriquez, "A computer model of normal conduction

in the human atria," Circ. Res., vol. 87, pp. 25-36, 2000.

[37] G. Constantine, K. Shan, S.D. Flamm, and M.U. Sivananthan, "Role of

MRI in clinical cardiology," Lancet, vol. 363, pp. 2162-71, 2004.

[38] R.M. Botnar, and M.R. Makowski, "Cardiovascular magnetic resonance

imaging in small animals," Prog. Mol. Biol. Transl. Sci., vol. 105, pp.

227-61, 2012.

[39] P. Glover, and P. Mansfield, "Limits to magnetic resonance

microscopy," Rep. Prog. Phys., vol. 65, pp. 1489-1511, 2002.

[40] E. Spuentrup, K.M. Ruhl, R.M. Botnar, et al., "Molecular magnetic

resonance imaging of myocardial perfusion with EP-3600, a collagen-

specific contrast agent: initial feasibility study in a swine model,"

Circulation, vol. 119, pp. 1768-75, 2009.

[41] G.L. Kung, T.C. Nguyen, A. Itoh, et al., "The presence of two local

myocardial sheet populations confirmed by diffusion tensor MRI and

histological validation," J. Magn. Reson. Imaging, vol. 34, pp. 1080-91,

2011.

[42] A.P. Benson, O.V. Aslanidi, H. Zhang, and A.V. Holden, "The canine

virtual ventricular wall: a platform for dissecting pharmacological

effects on propagation and arrhythmogenesis," Prog. Biophys. Mol.

Biol., vol. 96, pp. 187-208, 2008.

[43] P.G. Batchelor, D. Atkinson, D.L. Hill, F. Calamante, and A. Connelly,

"Anisotropic noise propagation in diffusion tensor MRI sampling

schemes," Magn. Reson. Med., vol. 49, pp. 1143-51, 2003.

[44] V. Jacquemet, A. van Oosterom, J.-M. Vesin, and L. Kappenberger,

"Analysis of electrocardiograms during atrial fibrillation. A biophysical

model approach," IEEE Eng. Med. Biol. Mag., vol. 25, pp. 79-88, 2006.

[45] M. Ridler, D.M. McQueen, C.S. Peskin, and E. Vigmond, "Action

potential duration gradient protects the right atrium from fibrillating," In

Conf. Proc. 2006 IEEE Eng. Med. Biol. Soc., pp. 3978-81.

[46] O.V. Aslanidi, T.D. Butters, C.X. Ren, G. Ryecroft, and H. Zhang,

"Electrophysiological models for the heterogeneous canine atria:

computational platform for studying rapid atrial arrhythmias," In Conf.

Proc. 2011 IEEE Eng. Med. Biol. Soc., pp. 1693-96.

[47] J. Zhao, M.W. Krueger, G. Seemann, et al., "Myofiber orientation and

electrical activation in human and sheep atrial models," In Conf. Proc.

2012 IEEE Eng. Med. Biol. Soc., to be published.