Abstract
1528
Objectives The design of concise and fast PET/MR protocols that provide an efficient workflow in clinical routine is mandatory. Our aim was to evaluate the accuracy in lesion detection of a short PET/MR protocol in comparison to PET/CT.
Methods Twenty-four patients sent to our Nuclear Medicine Department for solid tumor staging underwent a dual-imaging protocol consisting of a PET/CT scan (Biograph, Siemens) followed by PET/MR scan (Signa, GE healthcare, with TOF capability). PET/CT scans were performed according to the standard clinical protocol (59±13 min after injection of 4 MBq/kg of 18F-FDG, head to mid-tight, 3 min/bed position, 7 beds; PET data reconstructed with a FORE+OSEM approach using 6 iterations and 80 subsets). Subsequently (111±22 min after injection), PET/MR was performed (2min/bed position, 5 beds, PET data reconstructed using a time-of-flight reconstruction with OSEM using 2 iterations and 28 subsets). MR sequences consisted in a Dixon sequence for attenuation correction purpose and only a thin section (5mm) axial T2-weighted fast recovery fast spin-echo (T2W FRFSE) with respiratory triggering (~2 min). The most representative hypermetabolic lesions (2 per organ, 5 maximum per patient according to PERCIST guidelines) were identified on PET/CT and PET/MR. The number, location, maximum standardized uptake value (SUVmax) and largest anatomical diameter (measured on CT and MR) of these lesions were recorded and compared between both modalities. Correlations and comparison between continuous variable were calculated using Spearman's rank coefficients and paired Wilcoxon tests, respectively.
Results Sixty-seven hypermetabolic lesions were identified both on PET/CT and PET/MR: lymph nodes (n=32), lung nodules (n=21), sus-mesocolic abdominal area lesions (n=7), musculoskeletal (n=4), breast (n=2), and cervix lesions (n=1). The corresponding anatomical lesion was measurable more often in PET/MR than in PET/CT: 81% (54/67) vs 73% (49/67), respectively. Six lesions (6/67) were measurable only on axial T2W FRFSE sequence and were related to 4 musculoskeletal lesions and 2 mediastinal LN, whereas only 2 mediastinal LN (2/67) were measurable only on CT, because of artifacts on T2W images. A high correlation between PET/CT and PET/MR regarding SUVmax (r=0.93, p < 0.0001) and anatomical size (r=0.98, p < 0.0001) was observed. SUVmax were 70±58 % higher in PET/MR than PET/CT, due to both the ~ 50 min delay between PET/MR and PET/CT, and the different generation of PET scanners (TOF vs non-TOF). Anatomical size of the lesions were slightly smaller (-4±21%) on T2W images than on CT (paired Wilcoxon test, p=0.004).
Conclusions This PET/MR short protocol, including axial thin-section T2W FRFSE and respiratory triggering, was feasible with high quality in short acquisition time (around 10 min). Lesion detection, location, as well as measurements in PET/MR involving T2W FRFSE sequences were comparable to those in PET/CT using CT without contrast. Thus, if the total target PET/MR scan duration is around 30 min for maintaining high patient throughput, 20 minutes are available for complementary organ specific MR exploration (brain, liver, head and neck or pelvic), making PET/MR a more comprehensive scanning procedure than PET/CT for patient management.