| Title: | Fully Automated, Quality-Controlled Cardiac Analysis From CMR: Validation and Large-Scale Application to Characterize Cardiac Function |
| Journal: | JACC Cardiovasc Imaging |
| Published: | 1 Mar 2020 |
| Pubmed: | https://pubmed.ncbi.nlm.nih.gov/31326477/ |
| DOI: | https://doi.org/10.1016/j.jcmg.2019.05.030 |
Publication 2223
WARNING: the interactive features of this website use CSS3, which your browser does not support. To use the full features of this website, please update your browser.
Abstract
We obtained reference standards measures for LV function, RV function and LV strain from short and long axis cine CMR. We trained networks to obtain these measures and then had experienced CMR cardiologists manually check these measures for their validity in a group of 700 cases. From these cases, we have a very high confidence that they are right, since they have been checked by clinicians.Deep learning (DL) algorithms can analyse cardiac magnetic resonance (CMR) images, but the presence of unanticipated analysis-errors currently excludes their use in clinical practice. We developed a pipeline for DL-based analysis of cine CMR, including quality control (QC) steps that prevent erroneous outputs. We show good agreement of the pipeline versus manual analysis, and high sensitivity of the QC steps in detecting erroneous results.
1 Author
- Ruijsink B et al.
1 Application
| Application ID | Title |
|---|---|
| 17806 | Motion-based diagnosis and characterisation of heart disease |
Enabling scientific discoveries that improve human health