The effect of misregistration between CT-attenuation and PET-emission images in 13N-ammonia myocardial PET/CT

Abstract

In 2-dimensional cardiac PET/CT, misregistration between the PET and CT images due to respiratory and cardiac motion causes tracer uptake to appear substantially reduced. The resolution and quality of the images have been considerably improved by the use of 3-dimensional (3D) PET acquisitions. In the current study, we investigated the impact that misregistration between PET and CT images has on myocardial 13N-ammonia uptake in images reconstructed with 3D ordered-subset expectation maximization combined with time-of-flight and point-spread-function modeling. Methods: Eight healthy volunteers (7 men and 1 woman; mean age ± SD, 53 ± 19 y) underwent 13N-ammonia cardiac PET/CT at rest. First, any misregistration between the PET and CT images was manually corrected to generate reference images. Then, the images were intentionally misregistered by shifting the PET images from the reference images by a degree of 1, 2, 3, 4, 5, 10, and 15 mm along both the x-axis (left) and the z-axis (cranial). For each degree of misregistration, the PET images were reconstructed using the CT-attenuation images. The left ventricular short-axis PET/CT images were divided into 4 segments: anterior wall, inferior wall, lateral wall, and septum. The erroneous decrease in myocardial uptake in basal, mid, and apical slices was visually graded using a 4-point scale (0 5 none, 1 5 mild, 2 5 moderate, and 3 5 severe). Wall-to-septum uptake ratios were evaluated for the anterior, inferior, and lateral walls in the basal, mid, and apical slices. Results: A statistically significant reduction in myocardial 13N-ammonia uptake in the anterior (P < 0.01) and lateral (P < 0.05) walls was observed when misregistration was 10 mm or more. The uptake ratios for the anterior, lateral, and inferior walls in the reference images were 1.00 ± 0.04, 0.96 ± 0.08, and 0.91 ± 0.03, respectively. The ratios for the anterior and lateral walls significantly decreased when misregistration exceeded 10 mm (anterior wall, 0.80 ± 0.06, P < 0.0001; lateral wall, 0.82 ± 0.07, P < 0.01), whereas the ratio for the inferior wall was relatively small at all 7 degrees of misregistration (0.86 ± 0.05 at 15-mm misregistration, P = 0.06). Conclusion: In PET/CT images reconstructed with 3D ordered-subset expectation maximization combined with time-of-flight and point-spreadfunction modeling, we found a statistically significant artifactual reduction in tracer uptake in heart regions overlapping lung when misregistration between PET and CT exceeded 10 mm.