Key result
Deep neural networks and graph matching improve congenital heart disease segmentation accuracy ~12% versus Seg-CNN.
Why the study?
Automatic segmentation of the whole heart and great vessels on CT images has been under-researched due to significant anatomical variations, and existing algorithms perform poorly in complex congenital heart disease.
Does a framework combining deep learning and graph matching improve whole heart and great vessel segmentation accuracy in congenital heart disease CT images compared to Seg-CNN?
Population
68 3D CT images covering 14 types of CHD
Comparison
Combined deep learning and graph matching framework vs state-of-the-art segmentation method in normal anatomy
Design
Algorithm development and validation study
Authors
Loading...
Hybrid DL-graph matching boosts CHD CT segmentation; hypothesis-generating for clinical workflows pending validation.
Does a framework combining deep learning and graph matching improve whole heart and great vessel segmentation accuracy in congenital heart disease CT images compared to Seg-CNN?
Absolute Event Rate: 78.3% vs 66.5%
p-value: p=<0.05
A novel framework combining deep learning and graph matching significantly improves the accuracy of whole heart and great vessel segmentation in complex congenital heart disease CT images compared to existing methods.
Yao et al. (2023) studied Congenital heart disease (n=68). Deep neural networks and graph matching framework vs. Seg-CNN was evaluated on Overall Dice score for whole heart and great vessel segmentation (p=<0.05). The proposed framework combining deep neural networks and graph matching achieved an overall Dice score of 78.3%, improving segmentation accuracy by 11.8% compared to Seg-CNN in congenital heart disease.
Synapse has enriched 2 closely related papers on similar clinical questions. Consider them for comparative context: