Session S62.2

Patient-Specific Cardiac Anatomical Models from Short-and Long-Axis MRI Slices

P Prior*, R Casero Canas, B Rodriguez, V Grau

Oxford University
Oxford, UK

Computational cardiac models have been shown useful for elucidating several important questions in cardiology. Clinical personalized cardiac models have great potential for improving cardiology patient management. This requires the generation of accurate, anatomically detailed cardiac meshes. Clinical imaging modalities like MRI can provide the anatomical information required to build the models; however the low out-of-plane resolution in clinical MRI scanners has been a limitation. By combining both short axis (SA) and long axis (LA) slices, this problem can be partially addressed. In this study, we report on a segmentation / post processing pipeline that converts MRI LA and SA slices of a patient into an anatomically realistic left ventricular mesh. A series of short- and long-axis slices of a human heart were acquired from a 1.5 T (63.7 MHz) clinical MR scanner using an SA True FISP cine protocol, with an in-plane resolution of 1.48 mm x 1.48 mm and out-of-plane resolution of 7 mm. After segmenting the SA and LA slices, the contour points were positioned in 3D space using the orientation information contained in the Dicom header, translated and rotated so that the main axes of the left ventricle (LV) corresponded to the coordinate frame axes, and transformed into spherical coordinates (?, ?, R). A thin-plate spline approximation with periodic boundary conditions was used to obtain an interpolated continuous surface R(?, ?). Additionally, the base of the LV was approximated by fitting a cubic spline to the end points of the LA contours. The process was repeated for both the epicardial and endocardial surfaces. Finally a volumetric mesh was generated by building tetrahedral between the epicardial and the endocardial points, exploiting the correspondence between the two surfaces in spherical coordinates. Results show the ability of the described algorithm to generate accurate, regular meshes which can be directly used to simulate cardiac activity using available simulation packages. Our aim is to use anatomically realistic meshes generated in this way to investigate cardiac electromechanical function in health and disease.

(Abstract Control Number: 213)