# Histological Validation of MR Diffusion Imaging for Myocardial Fiber Architecture Analysis

Walter G. O'Dell, Van Wedeen*, Timothy G. Reese* and Andrew D. McCulloch

Dept of Bioengineering, University of California San Diego
La Jolla, California, 92093-0412 USA

*Dept of Radiology, Massachusetts General Hospital
Charlestown MA, 02129




Introduction
Myocardial fiber and connective tissue architectures are fundamental determinants of regional and global cardiac mechanical function. Structural alterations in the myocardial tissue matrix are apparent within 90 minutes of mild ischemic injury[1] and gross changes in the fiber and collagen network occur over a much longer time scale in infarcted tissue. A method for measuring intercellular structure might therefore be a sensitive indicator of the severity, extent and time course of ischemic injury, while also providing a means to study myocardial structure/function relationships.
Imaging using diffusion-sensitive MRI has recently been demonstrated as a non-invasive method for computing fiber directions in various tissues[2,3]. Diffusion anisotropy has also been reported in myocardial tissue[4,5], however the relationship between the principal directions of water diffusion and the orientation of local myocardial fiber network has not been fully validated.

Methods
MR Diffusion Imaging: Isolated canine hearts were imaged using diffusion sensitive MRI and a spin-echo acquisition sequence. Six sets of biphasic gradient pulses, with diffusion sensitivity of B~300 s/mm^2, were used to compute the diffusion tensor at each imaging voxel. The principal directions of diffusion and their diffusion values were then calculated in each voxel and the voxel locations registered in 3D. An MR diffusion eigen-image was created by displaying the principal direction corresponding to (and scaled by) the largest eigenvalue, as shown in Figure 1.

The principal diffusion direction with the largest diffusion in the circumferential-longitudinal plane ({\it in-plane}) was assumed to represent the local fiber direction. The angle of the corresponding in-plane diffusion vector with respect to the local circumferential direction (and local epicardial surface normal) was then computed.

Fig. 1(Top) MRI diffusion principal eigen-value data set at a 24x24 image resolution (3x3x5mm voxel size). (Bottom) Picture of same heart noting locations of tissue sections and circumferential-radial (CR) and longitudinal-radial (LR) slices.

Histological analysis: Tissue blocks were harvested from 3 sites in the same heart used for the diffusion-sensitive imaging experiment. Transmural blocks of dimension ~1.5 cm^3 were excised, sectioned in orthogonal planes at 50--200micron thickness (see Figure 1) on a vibratome, transferred to a slide and viewed under low power (x10). Fiber angles were observed and recorded in sections cut orthogonal to the epicardial surface at ~1mm intervals.

 Apical Anterior Basal Anterior Basal Posterior
Fig. 2 Estimated myocardial fiber angles versus percent wall depth (from epicardium). Shown are measurements from histology of excised tissue blocks (unfilled circles), and MR diffusion-sensitive imaging (filled circles).

Results
Figure 2 shows plots of fiber angle versus percent wall depth for the histological and MR-diffusion data. The nearly linear variation of fiber angle with depth correlates with previous findings[6]. Linear regression lines were fit to each data set at each depth. The fiber angle intercepts at the epicardium and endocardium and fitted slopes are presented in Table 1.

 -45.6/-89.1 -66.8/-53.6 -56.3/-30.3 127/94 (histo/MR) apical ant. basal ant. basal post. epi int. endo int. slope
Table 1 Epi- and endo-cardial intercepts and slope of fiber angle versus wall depth from a linear regression fit. Data for both histological sectioning and MR diffusion imaging are given for 3 LV free-wall locations.

Discussion
A correlation is observed between the computed direction of largest in-plane diffusion and the histologically measured fiber angles in at least 2 of the 3 regions analyzed. The variance in the MR-based fiber angles is believed to arise from the uncertainty in the raw MR diffusion data, coupled with the relatively large MR voxel size compared to the observed transmural fiber angle variations. MR data from the extreme endo- and epi-cardial voxels are also likely to exhibit corruption due to signal from the fluid surrounding the heart. Fiber angle error at the intercepts is expected due to the uncertainty in delimiting the endo- and epi-cardial borders from the MR images. In the basal posterior site, where the slopes are least well correlated, there is a correspondance between the histologically measured angles and those from the MR analysis in the sub-epicardium where the distribution of fiber angle was highly non-linear.

Future work will investigate the relationship between the fiber laminal sheet structure[7] (perpendicular to the in-plane fiber direction) and the remaining two principal directions of diffusion. This knowledge is hoped to provide a formal framework for further investigations into the structural mechanisms of regional variations in myocardial function.

References
1. Zhao M et al,

J. Amer. College of Cardiology,
10:1322-34, 1995.
2. Wong EC, Cox RW and Song AW
Magnetic Resonance in Medicine,
34:139-143, 1995.
3. Hsu EW and Mori S,
Magentic Resonance in Medicine,
34:194--200, 1995.
4. Basser PJ, Mattiello J, and LeBihan D
Journal of Magnetic Resonance,
B103:247--254, 1994.
5. Reese TG, Weisskoff RM, Smith RN, Rosen BR
Dinsmore RE and Wedeen VJ
Proc of the SMR, 1:357, 1995.
6. Nielsen PMF, LeGrice IJ, Smaill BH, Hunter PJ
Am. J. Physiology, 260:H1365-H1378, 1991.
7. Legrice IJ, Smaill BH, Chai LZ, Edgar SG,
Gavin JB and Hunter PJ
Am. J. Physiology, 269:403-426, 1992.

Acknowledgements
This research was funded through NIH HL41603, NSF BES-9634974 and the Whittaker Foundation.