Key result
Automatic neural network for atrial fibrosis assessment outperforms manual segmentation with a 91% Dice score.
Why the study?
Adoption of atrial LGE has been hindered by nonstandardized, operator-dependent image processing, minimal validation, and limited access to transparent software platforms.
Does a fully automatic multilabel convolutional neural network accurately estimate atrial fibrosis from LGE-CMR scans compared to manual analysis?
Does a fully automatic multilabel convolutional neural network accurately estimate atrial fibrosis from LGE-CMR scans compared to manual analysis?
Absolute Event Rate: 91% vs 85%
A fully automatic multilabel convolutional neural network provides reproducible, operator-independent estimation of atrial fibrosis from LGE-CMR scans that is comparable to manual analysis.
No takes yet. Share an insight, caveat, or question.
Supports automated atrial fibrosis quantification on LGE-CMR; leaves open clinical adoption and outcome impact.
Razeghi et al. (2020) studied Atrial fibrosis (n=207). Fully automatic multilabel convolutional neural network pipeline vs. Manual fibrosis burden analysis was evaluated on Dice score for atrial segmentation. A fully automatic convolutional neural network for atrial fibrosis assessment achieved a 91% Dice score for atrial segmentation, outperforming the 85% manual interobserver agreement.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: