Pranav Bhagirath

45 Evaluation of state-of-the-art segmentation algorithms for left ventricle infarct from late Gadolinium enhancement MR images Table 1. Overview of previously published methods for scar quantification and segmentation. Reference Model n Algorithm Highlight LV Kim et al. (1999) Canine 26 2-SD Correlation of MRI enhancement with scar Amado et al. (2004) Animal 13 1-6 SD, FWHM FWHM correlates to histology Kolipaka et al. (2005) Human 23 2,3 SD Manual correction is necessary despite algorithm Positano et al. (2005) Human 15 Clustering Fast clustering algorithm Schmidt et al. (2007) Human 47 2-6 SD Grey-zone and core quantification Hennemuth et al. (2008) Human 21 EM fitting* Model based on scanner acquisition and reconstruction parameters Detsky et al. (2009) Human 15 Clustering* Clustering in feature space Tao et al. (2010) Human 20 Otsu thresholding Dice overlap on chronic myocardial infarction with 2-observer manual segmentation Flett et al. (2011) Human 60 2-6 SD, FWHM Inter- and intra-observer reproducibility Rajchl et al. (2014) Human 35 SD, FWHM, Max-flow Inter- and intra-observer reproducibility on 3D CMR Andreu et al. (2011) Human 12 50, 60, 70% FWHM 60% FWHM for good voltage correlation Lu et al. (2012) Human 10 Graph-cuts* Correlation with FWHM and manual segmentations on chronic myocardial infarction data Pop et al. (2013) Animal 9 Mixture model ex-vivo histology and high- resolution MRI LA Oakes et al. (2009) Human 81 2-4 SD LA fibrosis and correlation to recurrence Knowles et al. (2010) Human 7 MIP Necrosis and edema theory for reconnection, comparison with electroanatomical data Ravanelli et al. (2014) Human 10 SD, Skeletonisation* Comparison with electroanatomical data Karim et al. (2014) Human 15 Graph-cuts* Dice with 3-observer consensus delineation Harrison et al. (2014) Animal 16 2-6 SD ex-vivo histology infarct volume against MR Methods are listed in chronological order, type of data they were evaluated with and the algorithm for: left ventricle (LV) or left atrium (LA). Methods which report on a segmentation algorithm developed are marked with an asterix (*). MIP, maximum intensity projection.