Left Ventricle Wall Motion Analysis with Real-Time MRI Feature Tracking in Heart Failure Patients: A Pilot Study

Figures

• 

  Figure 1. Illustration of feature tracking algorithm for voxel-wise measurements of LV endocardial border motion along the radial and circumferential directions. In each time frame, a set of “line cuts” normal to the LV endocardia border were interpolated from the Cartesian image voxels and arranged to form a feature image for each voxel along the LV endocardial border. The voxel-wise tracking across time frames was performed by searching for the feature images which geometrically matched best. The radial displacement was given by the radius change of each voxel between two neighboring image time frames and the circumferential displacement by the angular change.

• 

  Figure 2. (A) Selected time frames of the retrospective cine and real-time images in a healthy volunteer and a heart failure patient. (B) Box plots for EDV, ESV, SV and EF measurements with retrospective cine and real-time imaging in the healthy volunteers and HF patients. The numbers above the boxes provide the P values from one-way ANOVA statistics and those below the boxes are the mean and standard deviation in each subject group.

• 

  Figure 3. Velocity measurements along the radial and circumferential directions in retrospective cine and real-time images from a healthy volunteer. (A) Motion trajectories of selected voxels along LV endocardial borders during systole and diastole within each cardiac cycle, respectively, in a basal and in an apical slice. (B) Radial and circumferential velocity measurements from the LV border trajectories in (A). The radial and circumferential velocity components were in-phase in the basal slice, but out-of-phase in the apical slice. Real-time measurements showed short-lived motion that varied from one cardiac cycle to another while retrospective measurements showed only long-lasting motion consistent over different cardiac cycles.

• 

  Figure 4. Peak velocity measurements with retrospective cine and real-time imaging in the healthy volunteers and HF patients. (A) Box plots for peak radial and circumferential velocity measurements in retrospective cine and real-time images. The numbers above the boxes provide p values from one-way ANOVA statistics and those below the boxes are the mean and standard deviation in each subject group. (B) Pearson correlation coefficients between peak velocity measurements and EF in retrospective cine and real-time images.

• 

  Figure 5. Examples of correlation analysis between radial and circumferential velocity measurements in retrospective cine and real-time images, respectively, from a healthy volunteer, from a HF patient with EF < 50%, and from a patient with HFpEF. The torsion correlation from real-time imaging was considerably different among three subjects while that from retrospective cine comparable.

• 

  Figure 6. Correlation analysis with retrospective cine and real-time images from 16 healthy volunteers, and 18 HF patients. (A) Box plots for the measurements of torsion correlation in the healthy volunteers (N = 16), HF patients with EF < 50% (N = 8) and patients with HFpEF (N = 10). The numbers above the boxes provide P values from one-way ANOVA statistics and those below the boxes are the mean and standard deviation in each subject group. (B) Scatter plots of EF against torsion correlation from retrospective cine and real-time images. The patients with HFpEF and the healthy volunteers were differentiated in the plots from real-time images while they were mixed in those from retrospective cine images.