An automatic contour extraction algorithm for short-axis cardiac magnetic resonance images

Area of Science:

• Cardiovascular Imaging
• Medical Image Analysis
• Biomedical Engineering

Background:

• Accurate segmentation of cardiac structures is crucial for diagnosing and monitoring cardiovascular diseases.
• Manual contour extraction from cardiac Magnetic Resonance Imaging (MRI) is time-consuming and subject to inter-observer variability.
• Automated methods are needed to improve efficiency and reproducibility in cardiac MRI analysis.

Purpose of the Study:

• To develop and evaluate an automatic process for extracting left ventricular (LV) contours from short-axis cardiac MR images.
• To minimize operator interaction, requiring only a single point within the LV for initialization.
• To assess the algorithm's performance on images with dark blood appearance.

Main Methods:

• The algorithm preprocesses cardiac MR images using smoothing and thresholding techniques.
• It generates edge images from segmented data to identify boundaries.
• Endocardial contours are extracted via a modified contour-following approach, and epicardial contours are approximated using radial search and Fast Fourier Transform (FFT) based smoothing.

Main Results:

• The automated method successfully extracts left ventricular endocardial and epicardial contours.
• The algorithm demonstrates robustness against moderate noise and minor gaps in the cardiac contours.
• Visual assessment by observers indicated success in processing images of reasonable quality.

Conclusions:

• The presented automatic contour extraction method offers a promising tool for left ventricular analysis in cardiac MRI.
• The algorithm requires further evaluation to fully determine its performance limitations and clinical applicability.
• This automated approach has the potential to enhance the efficiency and consistency of cardiac MRI interpretation.