Feasibility and results of a short whole-body PET/MR protocol for staging primary and secondary thoracic lesions from solid tumors

240 Objectives PET/CT is a pivotal tool in the staging of patients with primary or secondary lesions of the thorax. With PET/MRI, thorax exploration is classically described as suboptimal. We studied the feasibility and the results of a short PET/MR protocol including thin section T2W imaging and brain imaging in comparison to PET/CT. Methods Seventeen patients, in the setting of lung cancer staging or restaging (n=14/17) or breast cancer staging with thoracic lesions (n=3) underwent a dual-imaging protocol consisting of a PET/CT scan (Biograph, Siemens) followed by PET/MR scan (Signa, GE healthcare). PET/CT scans were performed according to the standard clinical protocols (63±11 min after injection of 4MBq/kg of 18F-FDG, head to mid tight, 3 min/bed position (n=7), PET data reconstructed with a FORE+OSEM, 6 iterations/80 subsets). Subsequently (108±14 min after injection), PET/MR was performed (2min/bed position (n=5), PET data reconstructed using a time-of-flight reconstruction with OSEM, 2 iterations/28 subsets). MR sequences included a Dixon sequence for attenuation correction and a thin section (5mm) axial T2-weighted fast recovery fast spin-echo (T2W FRFSE), with respiratory triggering (duration around 2 min). Eight out of the 17 patients had complementary MR brain imaging after the whole-body acquisition including a post contrast BRAVO sequence (Fast IR-prepared 3D gradient echo BRAin VOlume imaging) and a FLAIR (Fluid-attenuated inversion recovery) axial sequences. All thoracic lesions were identified visually in PET/CT and PET/MR. The number, location, SUVmax, and largest diameter of these lesions were recorded and compared between both modalities. Correlations and comparison between continuous variable were calculated using Spearman9s rank coefficient and paired Wilcoxon tests, respectively. Results Forty one mediastino-pulmonary hypermetabolic lesions, including 21 lung lesions (solid nodules n=18, ground glass nodule n=2, necrotic tumor n=1) and 20 lymph nodes (paratracheal n=8, hilar n=6, paraaortic n=6 and subcarinal n=1) were identified. The corresponding anatomical lesion was measurable in 83% (34/41) in PET/CT and in 83% (34/41) in PET/MR. All pulmonary lesions (n=21, mean CT size: 21.5+/-19 mm, min: 3mm, max: 69mm) seen on CT were also identified on thin section T2W FRFSE images, with an excellent correlation in lesion size (r=0.99) with no significant difference between MR and CT (p=0.2). SUVmax of lung lesions were 85±84 % 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). Two hilar LN (2/20) were measurable only on axial T2W FRFSE sequence because of the higher contrast than in CT between LN and bronchovascular elements, whereas 2 mediastinal LN (1 paratracheal and 1 para-aortic) were measurable only on CT because of artifacts on T2W images. High correlation (r=0.93, p Conclusions Whole-body PET/MR including brain imaging is feasible for staging thoracic malignancies, with scan duration less than 30 min. Lesion detection, location, as well as measurements in PET/MR involving a thin-section T2W sequence are comparable to those in PET/CT using CT without contrast. PET/MRI could be used instead of PET/CT if this approach becomes economically viable.