f Abstract— The ventricular myocardium has a structure of branching laminae through which course regularly orientated myofibers, an architecture important in excitation and contrac- tion. Quantifying this architecture is vital for understanding normal and disease states in the heart and for assessing their impact on electrical function. These data are also highly im- portant in the construction of scientifically and clinically useful computer models of cardiac electrical behavior. Detailed struc- tural information has previously been obtained from serial imaging. In this work we assess the potential for high- resolution (HR) MRI as a means to furnish useful myoarchitec- ture and compare and contrast this approach with the growing use of DT. Using rat hearts, we conclude that both approaches have strengths and weaknesses, however, HR-MRI may pro- vide a consistently more robust picture of the myoarchitecture in small hearts. I. INTRODUCTION The mammalian ventricles have a unique and specialized architecture consisting of a regular helical fiber-orientation which courses through a conserved and complex my- olaminar arrangement [1]. Due to the role of these structural features in electrophysiological and biomechanical function in both health and disease, their accurate measurement is important. Changes in fiber orientation and myolaminar sliding are thought to be the principle mechanisms of myo- cardial thickening in systole [2]. Fiber orientation has long been known to influence the spread of myocardial activation [3], and furthermore, laminar organization has recently been shown to substantially influence activation [4]. Myofiber and myolaminar structure are present throughout the myo- cardium (except myolaminae are absent in the immediate sub-epicardium [1]) and three principal orthogonal structural directions can be defined: (i) along the fiber axis; (ii) per- pendicular to the fiber axis in the laminar plane; and (iii) * This work was supported in part by grants from the Medical Research Council (G0701785, S. H. Gilbert). S.H. Gilbert is the corresponding author † O. Bernus and M.L. Trew contributed equally to this study. S.H. Gilbert and O. Bernus are with the Institute of Membrane and Sys- tems Biology, Faculty of Biological Sciences, Multidisciplinary Cardiovas- cular Research Centre, University of Leeds, Leeds LS2 9JT, United King- dom phone: +44-113-3431869; fax: +44-113-34342285; e-mail: {s.h.gilbert, o.bernus}@ leeds.ac.uk. S.H. Gilbert and O. Bernus are also with the Inserm U1045 - Centre de Recherche Cardio-Thoracique, L'Institut de rythmologie et modélisation cardiaque, Université Bordeaux Segalen, Centre Hospitalier Universitaire de Bordeaux, PTIB - campus Xavier Ar- nozan, Avenue du Haut Leveque, 33604 Pessac, France, ste- phen.gilbert@chu-bordeaux.fr, olivier.bernus@u-bordeaux2.fr. G.B. Sands, M.L Trew. I.J. LeGrice and B.H.Smaill are with the Auck- land Bioengineering Institute, The University of Auckland, UniServices House, Level 6, 70 Symonds Street, City Campus, Auckland 1010, New Zealand, Auckland, New Zealand. B.H. Smaill is also with the Department of Physiology, University of Auckland, Auckland, New Zealand. normal to the laminar plane. This structural arrangement is known as orthotropy [4]. Whole-heart computational model- ing requires detailed structural atlases. Ideally these would be from accurate high-throughput 3D imaging but current methods have considerable limitations. Diffusion Tensor - MRI (DT) has been widely used and has been validated for fiber and laminar measurements against 2D methods but it: (i) has limited spatial resolution [5]; (ii) has limited accuracy for laminae [6]; (iii) has not been validated against 3D methods; (iv) is SNR sensitive [5]; (v) the microstructural basis of the DT signal is controversial [7]; and, (vi) the in- fluence of the b-value has not been fully explored [7]. HR- MRI has high spatial-resolution, is applicable to the beating heart and it has been validated against 2D-histology [8, 9]. Structure tensor (ST) analysis is an image analysis method which derives a tensor from the distribution of gradient di- rections within the neighborhood of an image voxel [10, 11]. We hypothesized that ST analysis could be applied to HR- MRI images to quantify fiber and laminar orientation, and that myolaminar orientations from ST would be more accu- rate and reliable than those from DT, as the largest ST ei- genvalue corresponds to the sheet normal direction, whereas in DT it relates to the fiber direction. Computer models of cardiac arrhythmia are increasingly using high-resolution tissue structure images and experi- mental recordings from normal and pathological cardiac tissue for validation [12]. These models are used to probe the multi-scale and multi-dimensional structural basis for ar- rhythmias in a way not possible with experimental record- ings alone. Generally electrical activity is simulated using cell membrane models and tissue property parameter sets. Early computational models used histologically determined tissue geometries and fiber orientations or geometries from low resolution MRI [13] and recent models have used geom- etries and fiber orientations from DT [14], or high-resolution geometries from high-field MRI along fiber and laminar orientations from DT [12]. To the best of our knowledge ventricular fiber and laminar orientations determined direct- ly from HR-MRI have not previously been generated and used for computational electrophysiology simulations. Fur- thermore, ST and DT determined fiber and laminar orienta- tions have not been compared. II. METHODS A. Tissue Preparation Male Wistar rats (N = 5) weighing 200–220 g were eu- thanized in accordance with the UK Home Office Animals A Framework for Myoarchitecture Analysis of High Resolution Cardiac MRI and Comparison with Diffusion Tensor MRI* Stephen H. Gilbert, Gregory B. Sands, Ian J. LeGrice, Bruce H. Smaill, Olivier Bernus† and Mark L. Trew† 34th Annual International Conference of the IEEE EMBS San Diego, California USA, 28 August - 1 September, 2012 4063 978-1-4577-1787-1/12/$26.00 ©2012 IEEE