MOTIVATION: Atrial fibrillation (AF) is a cardiac arrhythmia characterized by a fast and irregular electrical activity in the atria. Highly prevalent (1-5%) and associated with a doubling of mortality and substantial morbidity, AF represents a high socio-economic burden [1]. The unsatisfactory efficiency of AF treatments demands a better understanding of the disease. AF is a complex disease, usually resulting from the interplay of several factors. One of the main factors is structural remodelling, which contributes to the accommodation and progress (paroxysmal to persistent to permanent) of AF itself [2]. One of the less studied structural changes is the increase in fat tissue. Available studies have only analysed the epicardium at relatively low resolution [3]. In this study, we analyse the accumulation of fat tissue in persistent AF, in the 3D structure of the goat left atrial wall (epicardium to endocardium) at unprecedented resolution. METHODS Tissue samples: In open chest experiments, left atrial wall tissue samples were excised maintaining tissue-pretension using rigid clip-on frames from two goat groups: persistent AF (5 goats) and control (6 goats). Persistent AF goats had been under AF for a period of 7 to 12 months. The collected tissue was stored in Karnovky’s fixative solution (2% paraformaldehyde, 2.5% glutaraldehyde in 0.1M sodium cacodylate buffer). Magnetic resonance imaging (MRI): Fixed tissue samples were preconditioned in gadolinium-containing (2mM) cacodylate buffer and then stabilised in an MRI tube using low melting 1% agar (Cambrex, Busieve GTG agarose). MRI scans were performed using a vertical-bore, 11.7 Tesla MR system with a Bruker console, a 40mm quadrature-driven birdcage coil (Rapid Biomedical, Würzburg, Germany), and a 3D fast gradient echo sequence: TE/TR 1.8/15ms, 15° pulse, field of view 40mm×40mm×40mm, matrix size 512×512×512, voxel size 78µm×78µm×78µm, and 10 averages. Image processing: MRI images were aligned in an epi- to endocardium view using a 3D rigid transformation. Thresholding and morphological operations were used to first correct the MRI background inhomogeneities, then segment tissue from the background, and finally segment fat from other tissue types. Validation of the fat segmentation was done using whole-mount pictures of tissue samples after excision and trichrome-stained histological slices (for details of histology see [4]). RESULTS and CONCLUSIONS We developed a method to successfully segment fat tissue in MRI data of atrial tissue samples. We are currently quantifying and comparing the accumulation of fat tissue in persistent AF versus control. Preliminary results suggest a larger volume of fat in AF then in control; and its preferential accumulation in near-epicardial layers. We will present the final analysis at the IUPS 2013 congress.