Related Experiment Videos
Automated left ventricular segmentation in cardiac MRI.
Amol Pednekar1, Uday Kurkure, Raja Muthupillai
1MR Clinical Science Group, Philips Medical Systems North America, Bothell, WA 98021, USA. amol.pednekar@philips.com
IEEE Transactions on Bio-Medical Engineering
|July 13, 2006
Summary
This study introduces an automated method for left ventricular (LV) myocardial boundary extraction. The novel approach demonstrates accuracy comparable to experienced radiologists, improving efficiency in cardiac imaging analysis.
Area of Science:
- Medical imaging
- Cardiology
- Image analysis
Background:
- Accurate delineation of the left ventricle (LV) myocardium is crucial for cardiac function assessment.
- Manual segmentation of myocardial boundaries is time-consuming and subject to inter-observer variability.
Purpose of the Study:
- To develop and validate an automated method for left ventricular myocardial boundary extraction.
- To compare the performance of the automated method against experienced radiologists.
Main Methods:
- Automatic localization of the LV using a motion map and expectation maximization algorithm.
- Segmentation of the myocardial region via an intensity-based fuzzy affinity map.
- Extraction of myocardial contours using dynamic programming for cost minimization.
Main Results:
- The automated algorithm achieved consistent mean bias of 7% when compared to experienced radiologists.
- The limits of agreement for the automated method were comparable to inter-observer variability in manual segmentation.
- Demonstrated feasibility of automated myocardial boundary extraction.
Conclusions:
- The developed automated method provides a reliable and efficient alternative for left ventricular myocardial boundary extraction.
- This technique has the potential to reduce variability and improve the consistency of cardiac image analysis.
- Further validation in diverse clinical settings is warranted.