Related Experiment Video
Updated: Dec 14, 2025
Author Spotlight: Bridging Gaps in Anatomy and Establishing a Foundation for Algorithmic Studies
Published on: December 15, 2023
Automated Segmentation of Cardiac Chambers from Cine Cardiac MRI Using an Adversarial Network Architecture
Roshan Reddy Upendra1, Shusil Dangi1, Cristian A Linte1,2
1Center for Imaging Science, Rochester Institute of Technology, Rochester, NY, USA.
Insights
This study compares U-Net and SegAN models for segmenting cardiac structures in cine cardiac magnetic resonance imaging (CMRI). SegAN models demonstrated improved performance in segmenting left and right ventricle blood-pools for better cardiac function analysis.
Area of Science:
- Cardiovascular Imaging
- Medical Image Analysis
- Artificial Intelligence in Medicine
Background:
- Cine cardiac magnetic resonance imaging (CMRI) is essential for cardiac function analysis, offering high spatio-temporal resolution.
- Accurate segmentation of cardiac structures (left ventricle blood-pool, myocardium, right ventricle blood-pool) is critical for diagnosing and monitoring heart conditions.
- U-Net based models represent the current state-of-the-art in medical image segmentation.
Purpose of the Study:
- To compare the performance of stand-alone U-Net models against U-Net models integrated within the SegAN framework.
- To evaluate segmentation accuracy for left ventricle blood-pool, myocardium, and right ventricle blood-pool.
- To determine the impact of SegAN's multi-scale loss function on cardiac image segmentation compared to single-scale U-Net models.
Main Methods:
- Utilized the 2017 ACDC segmentation challenge dataset comprising CMRI scans.
- Implemented and trained stand-alone U-Net models for cardiac structure segmentation.
- Implemented and trained U-Net models within the SegAN adversarial network framework, incorporating a multi-scale loss function.
- Quantified segmentation performance using mean Dice scores for each cardiac structure.
Main Results:
- Stand-alone U-Net models achieved mean Dice scores of 89.03% (LV blood-pool), 89.32% (myocardium), and 88.71% (RV blood-pool).
- U-Net models within the SegAN framework achieved mean Dice scores of 91.31% (LV blood-pool), 88.68% (myocardium), and 90.93% (RV blood-pool).
- The SegAN framework demonstrated superior performance for left ventricle blood-pool and right ventricle blood-pool segmentation compared to stand-alone U-Net.
Conclusions:
- The SegAN framework, leveraging a multi-scale loss function, enhances the segmentation accuracy of U-Net models for specific cardiac structures in CMRI.
- Improved segmentation of cardiac blood-pools and myocardium using SegAN can lead to more precise cardiac parameter computation.
- This approach holds potential for advancing cardiac disease diagnosis, therapy planning, and health monitoring through more accurate CMRI analysis.
Abstract:
Cine cardiac magnetic resonance imaging (CMRI), the current gold standard for cardiac function analysis, provides images with high spatio-temporal resolution. Computing clinical cardiac parameters like ventricular blood-pool volumes, ejection fraction and myocardial mass from these high resolution images is an important step in cardiac disease diagnosis, therapy planning and monitoring cardiac health. An accurate segmentation of left ventricle blood-pool, myocardium and right ventricle blood-pool is crucial for computing these clinical cardiac parameters. U-Net inspired models are the current state-of-the-art for medical image segmentation. SegAN, a novel adversarial network architecture with multi-scale loss function, has shown superior segmentation performance over U-Net models with single-scale loss function. In this paper, we compare the performance of stand-alone U-Net models and U-Net models in SegAN framework for segmentation of left ventricle blood-pool, myocardium and right ventricle blood-pool from the 2017 ACDC segmentation challenge dataset. The mean Dice scores achieved by training U-Net models was on the order of 89.03%, 89.32% and 88.71% for left ventricle blood-pool, myocardium and right ventricle blood-pool, respectively. The mean Dice scores achieved by training the U-Net models in SegAN framework are 91.31%, 88.68% and 90.93% for left ventricle blood-pool, myocardium and right ventricle blood-pool, respectively.
