Table 70 Publicly available data sets that have been used for training and testing for automated segmentation algorithms of the left and right ventricle
Data set | Conference/source | n | Segmented structure | Data description |
|---|---|---|---|---|
MICCAI-2009 [213] | MICCAI 2009 | 45 | Left ventricle | Single center, single vendor 5 sub-groups: healthy, hypertrophy, heart failure with infarction and heart failure without infarction Data hosted on: https://www.cardiacatlas.org/studies/sunnybrook-cardiac-data/ |
LVSC-2011 [214] | STACOM-2011 | 200 | Left ventricle | Multi center, multi-vendor Myocardial infarction |
RVSC-2012 [215] | MICCAI 2012 | 48 | Right ventricle | Single center Randomly selected clinical cases |
ACDC-2017 [198] | MICCAI 2017 | 150 | Left ventricle | Single center, 2 scanners, 1 vendor 5 sub-groups (Normal, post-myocardial infarction, dilated cardiomyopathy, hypertrophic cardiomyopathy, abnormal right ventricle) |
KAGGLE-2015 [216] | KAGGLE 2015s annual data science bow | 1100 | Left ventricle | Multi center, multi scanner Mix of patient and volunteer scans Only end-diastolic and end-systolic ground truth results provided. No gold standard segmentations available |
Multiple sources | Cardiac Atlas Project [217] | > 6500 | Left ventricle | Multi center, multi-vendor Asymptomatic subjects Data acquired with gradient echo cine acquisition Data hosted on: https://www.cardiacatlas.org/studies/mesa/ |