Automatic segmentation of cardiac MRI cines validated for long axis views

Area of Science:

• Cardiology
• Medical Imaging
• Biomedical Engineering

Background:

• Cardiac magnetic resonance imaging (CMR) segmentation is crucial for clinical assessment.
• Accurate delineation of endocardial and epicardial borders is essential for quantitative analysis.
• Manual segmentation is time-consuming and subject to inter-observer variability.

Purpose of the Study:

• To develop and validate an automatic algorithm for segmenting myocardial boundaries in cardiac MRI.
• To assess the accuracy and reliability of the proposed algorithm against gold standard measurements and manual segmentation.

Main Methods:

• An automatic algorithm was developed for segmenting endocardial and epicardial boundaries in long-axis CMR views.
• The algorithm was validated on a dataset of 126 patients (1008 traces).
• Performance was evaluated by comparing automated tracings to gold standard measurements and manual observer variability.

Main Results:

• The algorithm demonstrated high correlation between estimated clinical parameters and gold standard measurements.
• The segmentation error of the automatic method was not significantly different from the error between two manual observers.
• The tool provides reliable myocardial boundary segmentation in long-axis CMR views.

Conclusions:

• A robust automatic tool for myocardial boundary segmentation in cardiac MRI long-axis views has been developed.
• The algorithm achieves accuracy comparable to manual segmentation, offering a valuable clinical application.
• This automated approach can improve efficiency and consistency in cardiac MRI analysis.