Assessing the volume of right ventricle (RV) and its range of movement during a cardiac cycle currently requires a cardiologist to tediously track the RV boundary in the cardiac magnetic resonance (CMR) image sequence throughout a cardiac cycle. Even with simplified measurement such as tricuspid annulus plane systole excursion (TAPSE), manually identifying associated anatomical landmarks still makes the analysis procedure inefficient and inconsistent. To automate this procedure, we need an algorithm to automatically track the change in pixel coordinates for each such landmark as it moves from one image to the next during the cardiac cycle. However, existing registration-based cardiac motion estimation technique, which generally compute the motion for every pixel in the image, are quite complex and require an impractical amount of computation.

In order to make this calculation more feasible for widespread clinical use, our specific aim was to design a new non-rigid registration algorithm to estimate the motion of anatomical landmarks with high efficiency and the ability to handle highly nonlinear motion such as RV motion. This tracking system enabled us to efficiently explore new cardiac indices which better correspond to RV function. Finally, by combining the motion estimation from different image views, we created a linear RV volume estimator, which is more efficient and consistent than relying on manually drawn boundaries, and which potentially will have better accuracy than the inference from the single distance between apex and tricuspid annulus plane.

In this work, we implemented a preliminary automatic landmark tracking system and demonstrated it on a long-axis four-chamber-view CMR image sequence. Motion vectors are estimated for each pixel between two successive local sub-images. Then, by integrating the motion vectors through time, we obtain the trajectory of the landmarks defined in the first image frame.

The video shown above is tracking apex (top-right) and the lateral TA plane (bottom-left) (shown as two small green dots) from end-diastole landmark locations through the entire cardiac cycle. The image contrast has been enhanced for visual display.

This work was a collaborative effort with Dr. Vincent L. Sorrell in the College of Medicine at The University of Arizona.

Publications:

1. Srinivas L. Naik, Jeffrey J. Rodriguez, Nishant Kalra, and Vincent L. Sorrell, "Tricuspid Annular plane Systolic Excursion (TAPSE) Revisited Using CMR," Journal of Cardiovascular Magnetic Resonance, 2012, vol. 14 (Suppl 1), Feb. 1, 2012, p. 299. doi: 10.1186/1532-429X-14-S1-P299. Presented at the Society for Cardiovascular Magnetic Resonance (SCMR) 15th Annual Scientific Sessions, Orlando, FL, Feb. 2-5, 2012. [PDF ]
   
   
2. James L. Huang and Jeffrey J. Rodriguez "Non-Rigid Registration Using Gradient of Self-Similarity Response," Image and Vision Computing, vol. 32, no. 11, Nov 2014, pp. 825-834. [ PDF ]