Background: Dynamic motion of the heart due to cardiac and respiratory cycles, and rotation from varying patient positions between imaging modalities, can cause errors during cardiac image registration. This study used phantom, patient and animal models to assess and correct these errors. Methods and Results: Rotational errors were identified and corrected using different phantom orientations. ECG-gated fluoro images were aligned with similarly gated CT images in 9 patients, and accuracy assessed during atrial fibrillation (AF) and sinus rhythm. A tracking algorithm corrected errors due to respiration, where 4 independent observers compared 25 respiration sequences to an automated method. Following correction of these errors, target registration error was assessed. At 20 mm and 30 mm from the phantom model’s center point with an in-plane rotation of 8 degrees, measured error was 2.94 mm and 5.60 mm, respectively, and the main error identified. A priori method accurately predicted ECG location in only 38% (p=0.0003) of 313 R-R intervals in AF. A posteriori method accurately gated the ECG during AF and sinus rhythm in 97% and 98% of 375 beats evaluated, respectively (p=NS ). Tracking algorithm for ECG-gated motion compensation was identified as good or fair 96% of the time, with no difference between observers and automated method (chi-square=25; p=NS) . Target registration error in phantom and animal models was 1.75±1.03 mm and 0 to 0.5 mm, respectively. Conclusions: Errors during cardiac image registration can be identified and corrected. Cardiac image stabilization can be achieved using ECG gating and respiration.
展开▼
