Key result
A fully automatic learning-based method using deep learning and deformable models achieved an average Dice metric of 82.5% and Hausdorff distance of 7.85 mm for right ventricle segmentation.
Why the study?
Does a deep learning-based approach accurately segment the right ventricle from cardiac MRI?
Population
48 cardiac MRI datasets from the MICCAI 2012 RV Segmentation Challenge database
Comparison
Deep learning algorithms combined with… vs Existing techniques participated in the MICCAI…
Authors
Loading...
DL-based RV segmentation is technically feasible; leaves open external validation before clinical adoption.
Does a deep learning-based approach accurately segment the right ventricle from cardiac MRI?
Deep learning algorithms combined with deformable models provide accurate and robust automatic segmentation of the right ventricle from cardiac MRI, outperforming previous challenge techniques.
Avendi et al. (2017) studied Right ventricle segmentation from cardiac MRI (n=48). Deep learning algorithms (convolutional neural networks and stacked autoencoders) combined with deformable models vs. Existing techniques from the MICCAI 2012 challenge / ground truth contours was evaluated on Average Dice metric and average Hausdorff distance. A fully automatic learning-based method using deep learning and deformable models achieved an average Dice metric of 82.5% and Hausdorff distance of 7.85 mm for right ventricle segmentation.
Synapse has enriched 4 closely related papers on similar clinical questions. Consider them for comparative context: