Abstract
425
Objectives Textural indices (TI) measured from PET images might be useful biomarkers to characterize metabolic heterogeneity in tumor and tissues. PET images can currently be obtained either using PET/CT or PET/MR scanners. We investigated the changes in PET TI values in patients who underwent a PET/CT and a PET/MR scans and compared them to differences in SUV.
Methods 13 patients with various cancer types underwent a F18-FDG PET/CT (68±13 min post-injection, Siemens Biograph 6) followed by a PET/MR (114±15 min post-injection, GE SIGNA). PET/CT and PET/MR were reconstructed as performed in clinical routine (PET/CT: FORE+OSEM, 6 it, 80 subsets, no TOF, voxel size=4x4x3 mm3; PET/MR: OSEM 2 it/28 subsets, TOF, voxel size=3x3x3 mm3, segmentation-based attenuation correction using 3 compartments). For each patient, a 41.2±1.4 mL VOI was drawn over the healthy liver in the PET/CT and copied to the PET/MR with manual refinement of the VOI location to match the VOI position in the PET/CT. Each tumor lesion was segmented using a threshold value accounting for surrounding activity independently in the PET/CT and in the PET/MR scans. In each tumor VOI with volume > 2.2 mL, SUVmax, SUVpeak (sphere of 0.5 mL) and 6 TI (Entropy, Homogeneity, SRE, LRE, LGZE, HGZE) were calculated after resampling voxel intensities between 0 and 20 SUV using 64 discrete values. The relationships between PET/MR and PET/CT-based indices were investigated using Spearman correlation coefficients and percent differences. In the healthy liver, Bland-Altman plots were used to investigate the difference in SUV and TI between PET/CT and PET/MR.
Results In total, measurements in 13 liver VOI and 21 tumor VOI were analyzed (mean tumor VOI=16.2±18.8 mL). In the tumors, despite the ~55 min delay between PET/CT and PET/MR and the very different technologies of the two PET systems, TI measured from the PET/MR were highly correlated with those measured from the PET/CT, with Spearman r >0.82 (SRE) and up to 0.95 (Entropy), while r was 0.88 for SUVmax and 0.92 for SUVpeak. TI values were significantly different between PET/MR and PET/CT, with greater tumor heterogeneity on the PET/MR images (eg, mean PET/MR Entropy=2.46±0.35 vs mean PET/CT Entropy=2.17±0.33, p<0.05), possibly due to change in acquisition time and in spatial resolution in the reconstructed images. Yet, the mean percent differences in Entropy (-10%), Homogeneity (+17%), SRE (-1.1%) and LRE (4.7%) between PET/CT and PET/MR with respect to PET/MR were less than the mean percent difference in SUVmax (-35%) and SUVpeak (-32%), suggesting that these TI might be less dependent on the PET scanner and post-injection delay than SUVs. In the healthy liver, BA plots showed an excellent agreement of SUV and TI between PET/MR and PET/CT, with a mean difference of 0.70±0.32 for SUVpeak due to the delay between acquisitions, and 0.18±0.13 for Entropy for a mean Entropy varying between 0.93 and 1.22.
Conclusions TI were no more dependent on the type of PET scanners / imaging protocols than SUV measurements. All trends observed in TI measured in PET/CT were identical to those measured in PET/MR, suggesting that the results of studies regarding the usefulness of TI in PET/CT can translate to what would be observed in PET/MR.