Statistical shape model of the full heart built using Principal Component Analysis. The model consists of a non-manifold mesh of 13 cardiac components and coronary arteries. The mesh faces of each component are labeled using a prime number (>= 2); faces connecting two structures are labeled using the product of the two primes for unique identification. Additionally, an epiendo label indicates whether a vertex belongs to the epi- or endocardial surface. The model contains 41 modes of variation, accounting for 90 percent of variance in the training set. The norm of each mode vector is its associated eigenvalue (as opposed to 1). The model was generated from the end-diastolic shapes of 134 subjects. Five iterations of Procrustes alignment were applied. VTK XML Serial PolyData 6.49 MB