Respiratory Phase Matching in PET/CT Using Fast Spiral CT

Objectives: In a whole-body PET/CT study, patient motion during the CT and PET scans affects the spatial alignment of CT and PET images, sometimes leading to errors in image interpretation and PET attenuation correction. To address this problem, we have developed a method for automatically aligning PET and CT images in normal PET/CT protocols. The method uses a respiration-correlated waveform and has the same CT radiation exposure as in standard PET/CT. We demonstrate the improved alignment and the relationship between noise and distortion for different approaches to PET image reconstruction.

Methods: Our novel approach is called matched PET/CT and is to be compared with conventional static PET, amplitude-based gating into a single gate with retention of 35% of acquired data (HD Chest ), and phase-based gated PET. 25 oncology patients with known thoracic or liver lesions, 30 mm or less in size, were injected with ¹⁸F-FDG and scanned, 60 minutes later, in continuous bed motion mode on a Siemens mCT Flow PET/CT system. The protocol used a non-gated low-dose spiral CT scan during free breathing, followed by list-mode PET, also in free breathing. A respiration monitoring device collected waveform data during CT and PET, allowing measurement of respiratory phase and respiratory amplitude. Automated analysis of the waveform assigned a phase or amplitude to each CT slice. Matching images from the two modalities was based on determining, slice by slice, the PET phase or amplitude that most closely corresponded to CT. In the phase-based case, we used 8 or 16 PET gates. The latter resulted in less intra-frame blurring but more image noise. To measure the displacement between CT and PET in each case, we calculated the distance between the centers of gravity of 3D contours drawn on lesions on the respective PET and CT images. In the case of gated PET, we determined which gate corresponded to the smallest displacement value, a non-automated method that defined the so-called best gate for that lesion. Additionally, for each case, we evaluated the lesions’ maximum standardized uptake values (SUVmax) and characterized the image noise.

Results: In a preliminary analysis of 12 lesions with the phase-based method and 8 gates, average PET-CT displacements (and standard deviations) for matched, static, HD Chest, and the best gate were 4.2 ± 4.6, 4.9 ± 4.8, 6.4 ± 4.1 and 2.5 ± 3.5 mm . Lesion SUVmax were highest in HD Chest and gated images. When we compared images according to image noise, they ranked as follows from least to most noisy: static, HD Chest, matched based on 8 gates, 8 gates, matched based on 16 gates, and 16 gates. We evaluated three methods of interpolating between gates. Nearest neighbor interpolation resulted in distorted, noisy images. Bilinear interpolation reduced distortions but introduced a noise pattern that varied between slices. Convolution with a 1-2-1 kernel resulted in a comparatively low noise and a reduced level of image distortions.

Conclusion: The alignment of PET and CT can be improved with an automated method based on measuring respiratory waveforms during CT and PET. The CT dose was unchanged from conventional PET/CT. Research Support: