LV wall segmentation using the variational level set method (LSM) with additional shape constraint for oedema...

Area of Science:

• Cardiovascular Imaging
• Medical Image Analysis
• Computational Pathology

Background:

• Myocardial edema extent after infarction indicates the area-at-risk (AAR).
• AAR estimation is crucial for determining salvageable myocardium post-myocardial infarction (MI).
• Oedema imaging holds significant clinical potential for acute MI patient management.

Purpose of the Study:

• To present an automatic algorithm for left ventricle (LV) wall segmentation and edema quantification.
• To utilize a variational level set method (LSM) with shape constraints for T2-weighted CMR image segmentation.
• To incorporate myocardial wall shape information into the variational LSM formulation.

Main Methods:

• Developed an automatic algorithm using variational level set method (LSM) with shape constraints.
• Applied the algorithm to segment the left ventricle (LV) wall in T2-weighted cardiac magnetic resonance (CMR) images.
• Quantified myocardial edema and compared results against manual segmentation and expert analysis.

Main Results:

• Mean perpendicular distances between automatic and manual LV wall boundaries were 1-2 mm.
• Bland-Altman analysis showed no consistent bias in LV wall area compared to two investigators.
• Edema quantification correlated well with expert assessment, with an average error of 9.3%.

Conclusions:

• The proposed automatic algorithm provides accurate LV wall segmentation and edema quantification.
• The method demonstrates high precision and minimal bias compared to manual segmentation.
• This technique offers valuable clinical utility in managing acute myocardial infarction patients.