TOF reconstruction for improved quantitative accuracy in PET/CT studies with misalignment

Objectives PET/CT misalignment is nearly inevitable in patient studies due to the different PET and CT acquisition characteristics and patient physiological changes in and between the PET and CT scans. Such misalignment can introduce significant errors to lesion quantitation. Time-of-flight (TOF) acquisition provides additional information of the origin of each detected event and is observed to be able to reduce PET/CT misalignment artifacts as compared to non-TOF. In this work, in addition to making general observations, we investigate if TOF can improve lesion quantitation over non-TOF for small to moderate PET/CT misalignment.

Methods We developed a simulation package to generate TOF data with various TOF resolution. The attenuation maps used for attenuation correction (AC) were shifted by 4 mm to 20 mm relative to the emission map to simulate small to moderate PET/CT misalignment. Localized misalignment was also simulated by shifting lung lesions only to mimic simple lung lesion misalignment due to respiratory motion. For patient studies acquired on a digital PET/CT system with TOF resolution of 325 ps, we introduced misalignment by shifting the CT images up (from patient back to chest) by 5 mm and 20 mm. Images were reconstructed using both non-TOF and TOF. Lesion quantitation errors were assessed by comparing the reconstructed lesion maximum to that without misalignment.

Results In the simulation studies, improved TOF resolution led to reduced lesion quantitation error and less prominent ghost lesions introduced by PET/CT misalignment. For a 20 mm misalignment, a 10 mm lung lesion with value 1.5 (lung 0.5, soft tissue 1.0) had reduced reconstructed value at the original lesion location (due to under AC); yet a ghost lesion was artificially generated at the location of the lung corresponding to the lesion mass in the attenuation map (due to over AC). The value at the original lesion location was increased from 0.3 in non-TOF to 0.5, 1.0, and 1.3 in TOF reconstruction with TOF resolution of 640, 320, and 160 ps, respectively; while the ghost lesion value was reduced from 2.2 in non-TOF to 1.6, 1.2, and 0.8 in TOF reconstruction. With the same misalignment, a 16 mm lung lesion with value 2.5 had reconstructed value increased from 1.5 in non-TOF image to 1.8, 2.2, and 2.4 in TOF images with TOF resolution of 640, 320, and 160 ps at the original location of the lesion, respectively; while the value of the introduced ghost lesion was reduced from 1.8 in non-TOF to 1.5, 1.0, and 0.7 in TOF images. In patient studies, the 5 mm misalignment led to lesion quantitation errors as much as 23% in non-TOF reconstruction. The Table listed four typical lesions (L1 to L4). L1 and L2 were relatively large and high contrast lesions in the middle of the lungs. Small to moderate misalignment led to relatively small quantitative changes in both non-TOF and TOF reconstructions. L3 was in mediastinum. The lesion was shifted to lung region of the attenuation map due to the misalignment, leading to under AC of the lesion. L4 was at lung/back boundary of the patient. The introduced misalignment shifted the lesion to the patient back in attenuation map, leading to over AC of the lesion. By using TOF reconstruction, the lesion quantitation errors were significantly reduced as compared to non-TOF. For 20 mm misalignment, non-TOF reconstruction showed greater than 10% errors for all the lesions. TOF reconstructed showed less than 10% errors for L1 and L2, but greater than 10% for L3 and L4. For 5 mm misalignment, the errors were greater than 10% for L3 and L4 in non-TOF, but all the lesion errors were less than 10% using TOF reconstruction.

Conclusions Using TOF reconstruction significantly reduced lesion quantitation errors caused by PET/CT misalignment. At TOF resolution of 325 ps, lesion quantitation was reliable for small PET/CT misalignment (<10% error for 5 mm misalignment).