Key result
IVUS-U-Net++ deep learning accurately segments IVUS lumen and media-adventitia borders with a ~0.92 Jaccard measure.
Why the study?
Current automated IVUS segmentation relies on manual corrections, which is time-consuming and user-dependent.
Does the IVUS-U-Net++ deep learning model accurately segment lumen and media-adventitia borders in IVUS images compared to ground truth?
Population
1746 IVUS images from 18 patients
Comparison
IVUS-U-Net++ segmentation model vs ground truth
Design
Model development and validation study
Authors
Loading...
May facilitate automated IVUS analysis in research; leaves open prospective clinical validation.
Does the IVUS-U-Net++ deep learning model accurately segment lumen and media-adventitia borders in IVUS images compared to ground truth?
p-value: p=<.01
The proposed deep learning model IVUS-U-Net++ accurately segments lumen and media-adventitia borders in IVUS images, showing strong agreement with ground truth measurements.
Zhu et al. (2022) studied Atherosclerosis (n=18). IVUS-U-Net++ deep learning model vs. Ground truth (manual segmentation) was evaluated on Jaccard measure (JM) and Hausdorff distance (HD) for lumen and media-adventitia borders, and correlation of 12 clinical parameters (p=<.01). The IVUS-U-Net++ deep learning model accurately segmented lumen and media-adventitia borders in IVUS images, achieving Jaccard measures of 0.9080 and 0.9199, respectively.
Synapse has enriched 4 closely related papers on similar clinical questions. Consider them for comparative context: