Diffusion Tensor Phenomapping of Local Microenvironments in the Pressure Overloaded Human Heart

Motivation: DTI data in the heart are currently averaged to yield per-patient values that cannot detect local variations in the microstructural microenvironment. Goal(s): We aimed to develop a voxel-based phenomapping approach to cluster the voxels in the myocardium into distinct groups based on their microstructural properties. We then used the approach to compare subjects with aortic stenosis (AS) to age-matched controls (CTL). Approach: A population matrix of voxels was created using z-score normalization and evaluated with k-means clustering (k=4). Results: 4 distinct clusters of voxels were present, each with different microstructural properties. No significant differences were seen between the AS and CTL subjects. Impact: An approach to perform voxel-based phenomapping of DTI data in the heart was developed and used to classify the voxels in the myocardium into distinct microstructural clusters.