Related Experiment Video
Updated: May 12, 2026
06:59
Improved Registration of 3D CT Angiography with X-ray Fluoroscopy for Image Fusion During Transcatheter Aortic Valve Implantation
Published on: June 3, 2018
A probabilistic patch-based label fusion model for multi-atlas segmentation with registration refinement: application
Wenjia Bai1, Wenzhe Shi, Declan P O'Regan
1Biomedical Image Analysis Group, Department of Computing, Imperial College London, SW7 2RH London, UK.
IEEE Transactions on Medical Imaging
|April 10, 2013
Summary
This study introduces a novel multi-atlas method for cardiac MRI segmentation, improving accuracy in evaluating ventricular function for cardiovascular disease diagnosis. The approach enhances segmentation precision for left and right ventricular cavities and myocardium.
Area of Science:
- Cardiovascular Imaging and Diagnostics
- Medical Image Analysis
- Computational Anatomy
Background:
- Accurate evaluation of ventricular function is crucial for diagnosing cardiovascular diseases.
- Current methods for measuring left ventricular (LV) mass and volume rely on manual segmentation, which is time-consuming and subjective.
- Objective and automated methods are needed to improve the efficiency and reproducibility of cardiac image analysis.
Purpose of the Study:
- To propose and evaluate a novel multi-atlas method for cardiac magnetic resonance (MR) image segmentation.
- To enhance the accuracy of left ventricular (LV) and right ventricular (RV) cavity and myocardium segmentation.
- To provide a more objective and efficient tool for assessing ventricular function in clinical settings.
Main Methods:
- A multi-atlas approach for cardiac MR image segmentation was developed.
- A patch-based label fusion model was formulated within a Bayesian framework.
- Image registration accuracy was improved by incorporating label information, leading to enhanced segmentation accuracy.
Main Results:
- The proposed method achieved high segmentation accuracy on a dataset of 28 cardiac MR images.
- Average Dice overlap metrics were 0.92 for the LV cavity, 0.89 for the RV cavity, and 0.82 for the myocardium.
- The results demonstrate the method's capability to provide precise quantitative information for clinical diagnosis.
Conclusions:
- The novel multi-atlas segmentation method offers accurate and reproducible ventricular segmentation from cardiac MR images.
- This technique can significantly aid in the diagnosis and management of cardiovascular diseases by providing reliable ventricular function metrics.
- The proposed Bayesian framework and label-informed registration represent advancements in automated cardiac image analysis.
