Unbiased stratification of left ventricles

Area of Science:

• Cardiovascular Imaging and Analysis
• Computational Anatomy
• Biomedical Engineering

Background:

• Quantitative stratification of left ventricles (LV) is crucial for understanding heart structure-function relationships.
• Current methods lack unbiased, reference-less grouping and physioanatomically relevant alignment for robust cardiac atlas construction.

Purpose of the Study:

• To develop a novel algorithm for unbiased stratification and alignment of LV shapes.
• To enable the construction of a pathology-stratified cardiac atlas.
• To adapt clinical strategies for automated cardiac shape analysis.

Main Methods:

• Individual LV shape models (N=127) were oriented to an 'attitudinally consistent orientation' capturing morphological variations.
• Affinity propagation technique was employed for grouping LV shapes based on automatically identified inter-LV landmark distances.

Main Results:

• The proposed algorithm achieves unbiased stratification and physioanatomical alignment of LV shapes.
• The method is computationally efficient.
• The approach successfully groups LV shapes based on landmark distances.

Conclusions:

• This study presents a novel, efficient method for LV shape stratification and alignment.
• The developed technique can facilitate the creation of clinically relevant, pathology-stratified cardiac atlases.
• Linking inter-cluster variations to pathology could significantly advance cardiovascular research and clinical practice.