SPASM: A 3D-ASM for segmentation of sparse and arbitrarily oriented cardiac MRI data Hans C. van Assen a,1 , Mikhail G. Danilouchkine a,2 , Alejandro F. Frangi b , Sebastia ´n Orda ´s b , Jos J.M. Westenberg a , Johan H.C. Reiber a, * , Boudewijn P.F. Lelieveldt a a Division of Image Processing, Department of Radiology, Leiden University Medical Center, Albinusdreef 2, PO Box 9600, 2300 RC, Leiden, The Netherlands b Computational Imaging Laboratory, Department of Technology, Universitat Pompeu Fabra, Barcelona, Spain Received 30 March 2005; received in revised form 29 November 2005; accepted 7 December 2005 Available online 24 January 2006 Abstract A new technique (SPASM) based on a 3D-ASM is presented for automatic segmentation of cardiac MRI image data sets consisting of multiple planes with arbitrary orientations, and with large undersampled regions. Model landmark positions are updated in a two-stage iterative process. First, landmark positions close to intersections with images are updated. Second, the update information is propagated to the regions without image information, such that new locations for the whole set of the model landmarks are obtained. Feature point detection is performed by a fuzzy inference system, based on fuzzy C-means clustering. Model parameters were optimized on a computer cluster and the computational load distributed by grid computing. SPASM was applied to image data sets with an increasing sparsity (from 2 to 11 slices) comprising images with different orientations and stemming from different MRI acquisition protocols. Segmentation outcomes and calculated volumes were compared to manual segmentation on a dense short-axis data configuration in a 3D manner. For all data configurations, (sub-)pixel accuracy was achieved. Performance differences between data configurations were significantly different (p < 0.05) for SA data sets with less than 6 slices, but not clinically relevant (volume differences < 4 ml). Compar- ison to results from other 3D model-based methods showed that SPASM performs comparable to or better than these other methods, but SPASM uses considerably less image data. Sensitivity to initial model placement proved to be limited within a range of position pertur- bations of approximately 20 mm in all directions. Ó 2005 Elsevier B.V. All rights reserved. Keywords: SPASM; Active shape model; ASM; 3D; Sparse; MRI; CT; Segmentation; Cardiac; Short-axis; SA; Long-axis; LA; Radial 1. Introduction 1.1. Purpose Nowadays, cardiac MRI and CT are increasingly used for cardiac functional analysis in daily clinical practice. Both modalities yield dynamic 3D image data sets. With CT, images are acquired in an axial orientation and for car- diac analysis, usually short-axis views are reconstructed from the axial image data. With MRI, images can be acquired in any spatial orientation. Commonly used orien- tations are short axis (SA) views (van Rossum et al., 1987), and long-axis (LA) views (2-chamber and 4-chamber). 1361-8415/$ - see front matter Ó 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.media.2005.12.001 * Corresponding author. Tel.: +31715261117; fax: +31715266801. E-mail addresses: h.c.v.assen@tue.nl (H.C. van Assen), m.danilouchkine @erasmusmc.nl (M.G. Danilouchkine), alejandro.frangi@upf.edu (A.F. Frangi), sebastian.ordas@upf.edu (S. Orda ´s), j.j.m.westenberg@ lumc.nl (J.J.M. Westenberg), j.h.c.reiber@lumc.nl (J.H.C. Reiber), b.p.f.lelieveldt@lumc.nl (B.P.F. Lelieveldt). 1 Present address: Biomedical Image Analysis, Biomedical Engineering, Technical University Eindhoven, Eindhoven, The Netherlands. 2 Present address: Department of Cardiology, Erasmus Medical Center, Rotterdam, The Netherlands. www.elsevier.com/locate/media Medical Image Analysis 10 (2006) 286–303