Abstract 6114: A Newly Developed Method Computes Cardiac Torsion from 4-Dimensional Ultrasound Data
Left ventricular (LV) torsion is a sensitive index of cardiac performance. 2D methods to compute torsion require multiple short axis views and are subject to error due to through-plane myocardial motion. We developed a new and unique 4D echo based method for quantifying cardiac mechanics, including torsion. In 7 open-chest pigs (8–10kg), 7 sonomicrometry crystals were implanted into the LV myocardium: 3 crystals at the apex and 4 close to the base. Full volume 4D image loops were acquired at 24–38 volumes per heart cycle on a Philips ie33 system. Torsion was computed as the angle of rotation between the two planes defined by the two sets of crystals. For the 4D image data, polar coordinate images were entered into an envelope detection program in MatLab. The 3D positional displacement field modeled using B-splines was computed by a non-rigid registration method minimizing the sum of squared intensity differences between systole and diastole. To compare torsion derived from the registration algorithm to those from sonomicrometry, we visually located the crystal positions on the myocardium. Temporary LAD occlusion was accomplished on 5 pigs to alter LV mechanics. With LAD occlusion, average cardiac torsion was reduced and detected both by sonomicrometry (5.8 ± 0.5 to 4.5 ± 0.5) and the 4D method (6.5° ± 0.5 to 4.2° ± 0.5) with good correlation between both methods We computed correlation coefficients (CC) between algorithm and sonomicrometry-derived torsions, which were averaged across all volumes. The average CC was 0:93. This new method can be used to accurately compute LV torsion, in 4D space at baseline and the changes induced by LAD occlusion.