Related Experiment Video
Updated: Mar 7, 2026
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
43.7K
Automatic segmentation of the right ventricle from cardiac MRI using a learning-based approach
Michael R Avendi1,2,3, Arash Kheradvar1,2, Hamid Jafarkhani3
1The Edwards Lifesciences Center for Advanced Cardiovascular Technology, University of California, Irvine, California, USA.
Magnetic Resonance in Medicine
|February 17, 2017
Summary
This study introduces an automatic deep learning method for segmenting the right ventricle (RV) in cardiac MRI scans. The technique achieves high accuracy, outperforming existing methods in RV segmentation challenges.
Area of Science:
- Medical Imaging
- Cardiovascular Research
- Artificial Intelligence in Medicine
Background:
- Accurate segmentation of the right ventricle (RV) in cardiac magnetic resonance imaging (MRI) is crucial for diagnosing cardiovascular diseases.
- Manual segmentation is time-consuming and prone to inter-observer variability.
- Automated methods are needed to improve efficiency and reproducibility.
Purpose of the Study:
- To develop and validate a fully automatic, learning-based method for segmenting the right ventricle (RV) from cardiac MRI.
- To enhance the accuracy and robustness of RV segmentation using deep learning and deformable models.
Main Methods:
- Utilized deep learning algorithms, specifically convolutional neural networks and stacked autoencoders, for initial RV chamber detection and segmentation.
- Integrated initial segmentation results with deformable models to refine accuracy and robustness.
- Trained the algorithm on 16 cardiac MRI datasets from the MICCAI 2012 RV Segmentation Challenge and validated on 32 additional subjects.
Main Results:
- Achieved an average Dice metric of 82.5% and an average Hausdorff distance of 7.85 mm for RV segmentation.
- Demonstrated high correlation and agreement with ground truth for key functional parameters: end-diastolic volume (0.98), end-systolic volume (0.99), and ejection fraction (0.93).
Conclusions:
- Deep learning algorithms are effective for automated RV segmentation in cardiac MRI.
- The proposed method's quantitative metrics surpassed those of other techniques in the MICCAI 2012 challenge.
- This automated approach offers a promising tool for clinical cardiovascular assessment.
