Cine DENSE strain imaging of the right ventricle: improved methods and initial experience in heart failure

Building upon prior methods for LV analysis [1, 2], we developed improved algorithms for semi-automatically analyzing RV strain. The first stage of RV analysis is anatomical delineation, locating DENSE observations attributable to underlying RV tissue. The algorithm requires user definition of RV anatomy on a single long-axis frame using a line contour (Figure 1a, b, c). Next, using a phase-quality-guided path-finding algorithm for phase unwrapping and using spatial interpolation with radial basis functions (RBF), the user-defined contour is propagated to all other cardiac phases through a process termed motion guided segmentation (MGS) [1]. With MGS, we project the manually defined RV contour to other cardiac phases using RBF interpolation of the displacement field. Lastly, material point trajectories are fit through time via a high order polynomial function. The final phase of RV analysis is strain estimation. Given the thin structure of the RV, we used a 1D contour strain, as illustrated in Figure 1. These new algorithms were applied to long-axis cine DENSE images acquired from 18 subjects without heart disease and 16 subjects with heart failure. Four data sets from HF patients were excluded due to aliasing artifacts. The RV was divided into two segments, namely the RV free wall and the septum.

The new DENSE analysis methods successfully propagated RV contours to sequential cardiac frames and computed RV contour strains. Using these methods, cine DENSE detected abnormal RV strain in HF patients compared to subjects without heart disease.

NIH K23 grant HL094761 and American Heart Association Grant-in-Aid 12GRNT12050301.