Alexander James Wilson1, Kevin M Moulin2, Gregory B Sands3, and Daniel B Ennis2
1Radiology, Stanford University, Palo Alto, CA, United States, 2Radiology, Stanford University, Stanford, CA, United States, 3Auckland Bioengineering Institute, University of Auckland, Auckland, New Zealand
Synopsis
Aiming to improve noninvasive
assessment of tissue microstructure, a physics-based simulation of diffusion
tensor imaging (DTI) was used to compare helix angles between DTI and structure
tensor (ST) analyses of a confocal myocardial image stack. Rodent myocardium
was imaged using extended volume confocal microscopy producing a high-resolution
image volume, which was segmented into intracellular and extracellular
compartments. The image volume was divided into voxel-blocks, which served as
the DTI simulation voxels. The DTI acquisition showed good agreement with both
the ST and manually measured helix angles.
This proof-of-concept work demonstrates the feasibility of direct
comparison of confocal images with DTI.
Introduction
Myocardial biopsy and histology are the
current reference standard to diagnose conditions that afflict the
microstructure of the myocardium, such as myocarditis and transplant
rejection. These invasive procedures
have an associated risk of damage to the heart and lungs. Noninvasive measurement of cardiac
microstructure would avoid these risks.
Diffusion tensor MRI probes the
Brownian motion of water molecules to gain information about micro-structures
within the imaging voxel. Diffusion weighted acquisitions are performed along
different directions, and the diffusion tensor model can be fit to the image
data. Cardiac diffusion tensor imaging (cDTI) is a technique for studying
cardiac microstructure, for which the primary eigenvector aligns with the myofiber
direction (Refs 5-7).
Methods such as tissue clearing and
extended volume confocal microscopy allow for image acquisition of millimeter-scale
volumes with sub-micron spatial resolution. High resolution confocal images are
amenable to structure tensor (ST) analysis, which can also be used to assess
myofiber direction (Ref 4).
Herein we applied an existing
physics-based diffusion simulation framework (Ref 1) to a myocardial image
volume and compared the results against the image-based ST analysis. We demonstrate
feasibility of direct comparison between confocal images of myocardium with
cDTI.Methods
Figure 1 provides an overview
of the study, and shows histology and segmented images. A rodent heart was excised and stained with
picrosirius red. A transmural block of the left ventricular free wall was cut, dehydrated,
embedded in resin, and mounted on a precision stage. Extended-volume confocal
microscopy (Ref 2) produced the image volume (3952×988×247 microns, at 1 micron
voxel size). The ilastik
(v 1.3.3, Ref 3) pixel classification software was used to segment the image
volume into myocyte, extracellular matrix and blood vessel. The segmented image was then divided into 247×247×247
micron voxel-blocks, producing 64 voxel-blocks (Figure 2).
For each voxel-block
segmented, an intra/extra-cellular mask was generated to perform particle-based
and MR diffusion simulation (Ref 1). A total
of 1000 molecules per voxel-block were simulated with coefficients of diffusion
of 3×10-3 mm²/s
and 2×10-3 mm²/s, dT=10µs for the extracellular and intracellular
compartments respectively with no permeability. The diffusion MR sequence simulated corresponded
to a motion-compensated cDTI sequence with a b-value of 1000 s/mm2,
12 diffusion encoding directions and with 4 averages. A diffusion tensor was fit to the signals and
the primary, eigenvector was calculated.
The ST was estimated for each of the 64
voxel-blocks by calculating the gradient of intensity along each of the
confocal image directions (Figure 2), and these
derivatives were used to construct the ST. The third eigenvector of the ST has
the least change in contrast along its direction, and aligns with the myofiber
direction (Ref 4).
The helix angle (HA) was obtained by
projecting the myofiber orientation onto the circumferential-longitudinal
plane, and calculating the angle between that projection and the
circumferential direction.Results
The diffusion tensor simulation produced
a transmural helix angle of 49° ± 17° at the epicardium, -6° ± 4° at
the midwall, and -59° ± 17° at the
endocardium. This transition was smoothly
varying as it progressed through the voxel-blocks from epicardium to
endocardium (Figure 3).
The structure tensor analysis produced
a helix angle of -55° ± 1° at the epicardium, -13° ± 30° at the midwall, to -48°
± 1° at the endocardium. The expected
helix angle trend was observed in three of the voxel-block rows, with the top
voxel-block row deviating from this trend (Figure 4).
Compared with the manually measured
helix angle, both the diffusion tensor and ST approaches over-estimated of the HA
(Figure 5). The transmural gradient of the change in HA
for both the diffusion tensor and ST approaches matched the manually measured HA.Discussion
We demonstrated that a physics-based
simulation of cDTI acquisition was able to produce realistic myocardial HA that
aligned with histological images of the myocardium, as well as the feasibility
of direct comparison between confocal images and cDTI.
At the voxel-block level, the cDTI
performed similarly to the structure tensor analysis in most instances. The cDTI simulation had similar performance
to the image-based structure tensor analysis.
This study demonstrates the feasibility of using physics-based DTI
simulation to tune acquisition parameters to maximize sensitivity for a given
type of microstructural feature or histology specimen. While this study focused
on HA, acquisition strategies could in theory be developed to probe cellular
size to investigate myocyte hypertrophy, or the diffusion between blood vessel
and extracellular matrix as a means to probe microvascular dysfunction. This framework has the potential to allow pathologists
to try different diffusion sequences and parameters needed to extract relevant
microstructural information for the pathology specimen of interest.
One of the limitations of the
segmentation method used for this study, was that the boundaries of individual myocytes
were not distinctly defined, which would lead to an over-estimation of the
diffusion between the intracellular and extracellular compartments. In future work we aim to further compare real
and simulated cDTI acquisitions with both cDTI and histological imaging from the
same specimen.Conclusion
Using a physics-based simulation, we
were able to demonstrated that cDTI and histology can be directly compared,
allowing for future radiology-pathology studies.Acknowledgements
We acknowledge funding from the American Heart Association.References
1.
Moulin
K, Viallon M, Maforo NG, Mazzoli V, Croisille P, Ennis DB. Sensitivity of Cardiac Diffusion Encoding
Waveforms to Myocardial Extracellular Volume: A simulation study. ISMRM
28th Annual Meeting, Paris, France {Virtual Conference}, 2020.
2.
Sands,
G.B., Gerneke, D.A., Hooks, D.A., Green, C.R., Smaill, B.H. and Legrice, I.J.,
2005. Automated imaging of extended tissue volumes using confocal microscopy. Microscopy
research and technique, 67(5), pp.227-239.
3.
Berg,
S., Kutra, D., Kroeger, T., Straehle, C.N., Kausler, B.X., Haubold, C.,
Schiegg, M., Ales, J., Beier, T., Rudy, M. and Eren, K., 2019. Ilastik:
interactive machine learning for (bio) image analysis. Nature Methods,
pp.1-7.
4.
Bernus,
O., Radjenovic, A., Trew, M.L., LeGrice, I.J., Sands, G.B., Magee, D.R.,
Smaill, B.H. and Gilbert, S.H., 2015. Comparison of diffusion tensor imaging by
cardiovascular magnetic resonance and gadolinium enhanced 3D image intensity
approaches to investigation of structural anisotropy in explanted rat hearts. Journal
of Cardiovascular Magnetic Resonance, 17(1), p.31.
5.
Scollan, D.F., Holmes, A., Winslow, R. and
Forder, J., 1998. Histological validation of myocardial microstructure obtained
from diffusion tensor magnetic resonance imaging. American Journal of
Physiology-Heart and Circulatory Physiology, 275(6), pp.H2308-H2318.
6.
Holmes, A. Alexander, D. F. Scollan, and Raimond
L. Winslow. "Direct histological validation of diffusion tensor MRI in
formaldehyde‐fixed myocardium." Magnetic Resonance in Medicine: An
Official Journal of the International Society for Magnetic Resonance in
Medicine 44.1 (2000): 157-161.
7.
Kung, G.L., Nguyen, T.C., Itoh, A., Skare, S.,
Ingels Jr, N.B., Miller, D.C. and Ennis, D.B., 2011. The presence of two local
myocardial sheet populations confirmed by diffusion tensor MRI and histological
validation. Journal of Magnetic Resonance Imaging, 34(5),
pp.1080-1091.