Investigation of respiration motion of the heart based on semi-automated segmentation and modeling of respiratory-gated CT data

Joyoni Dey, IEEE
Tinsu Pan, University of Texas MD Anderson Cancer Center
Mark Smyczynski, University of Massachusetts Medical School
Hendrik Pretorius, IEEE
David Choi, University of Massachusetts Medical School
Michael A. King, IEEE


One of the factors limiting the diagnostic accuracy of cardiac SPECT perfusion imaging is the respiratory motion of the heart. Several authors have investigated the motion of heart due to respiration. In this work we have 4D-CT data for 7 patients, consisting of 10 respiration gated non-contrast CT datasets covering the heart region for each patient. We perform a segmentation and registration of the heart datasets in sequence to determine the gross rigid-body motion of the heart due to respiration. For each patient, we segment the heart with a prior shape with an initial pose on one coronal slice of one of the respiration stages, and then the algorithm tracks the object through the other coronal slices. The segmentation results for first stage of respiration are used to initiate the segmentation of the heart at second stage, and so on for the other stages of respiration. A 6-parameter rigid-body registration of the first stage of respiration to the 9 consequent stages estimates the gross motion of the heart. The results of tracking heart motion for the 7 patients indicate a SI axis translation with an (absolute) range of 2.6 to 10.7 mm and mean of 5.7 mm, and standard deviation of 3.7mm, during expiration. Mean rotations of 3.5 deg about the AP-axis, and 1.2 deg about the RL axis were also observed. © 2005 IEEE.