Abstract
1980
Objectives The replacement of photomultiplier tube (PMT) PET detector with digital photon counting (DPC) PET detector is a technological leap. While 4mm OSEM has been traditionally used for clinical wholebody PET reconstruction, integrating point-spread-function (PSF) onto high spatial resolution DPC PET to enable high-definition (HD, 2mm in voxel length) and ultra-high definition (UHD, 1mm in voxel length) PET reconstructions is a promising approach for improving lesion detectability. This study evaluates HD and UHD PET reconstruction on overall 18F-FDG PET image quality and lesion detectability compared to the standard definition (SD, 4mm OSEM without PSF) PET reconstruction.
Methods Datasets from 90s/bed wholebody scans of 30 clinical patients (75±17kg; 1.7±0.1m; 13±0.5mCi FDG) were utilized from on ongoing clinical trial acquired on a DPC PET/CT system (Vereos). Listmode acquired PET data were reconstructed in HD (3 iteration 13 subset) and UHD (3 iteration 9 subset) using 3D OSEM+TOF with PSF enabled, and compared to the default SD PET (3 iteration 15 subset) with no PSF. NEMA IEC Body phantom with hot spheres was used for evaluation and validation. Blinded image review was performed by experienced readers. Quantitative assessment in SUV, recovery coefficient (RC) and background noise were performed.
Results Lesion contrast is significantly improved on HD PET and UHD PET while the liver background remains equivalent compared to default SD PET. The 1mm UHD PET resulted in best in class image quality with the least partial volume effect and significantly improved contrast of lesion essential for small lesion detectability. 2mm HD PET generated great image quality with less noise contribution than 1mm PET for larger size patient (BMI > 25). Blinded image reviews scored higher for HD and UHD PET in regard to confidence in lesion detection than the default 4mm SD PET (p<0.01). 68 lesions are included in the evaluation, 14 of them of small size were only confidently detected on the HD and UHD reconstructed PET images. In correlation, on the phantom experiments, the smallest 10mm sphere with the lowest contrast (LBR=2) on phantom was clearly identified on the 1mm UHD PET. Quantitatively, more accurate SUV data were found on HD and UHD than on SD. It revealed RCs of 0.78-1.15 for UHD PET reconstruction and 0.70-1.06 for HD PET reconstruction compared to 0.58-0.98 for SD PET across spheres from 10mm to 37mm.
Conclusions This study demonstrated that High-Definition and Ultra-High-Definition PET reconstructions provide substantially improved image quality and diagnostic confidence for lesion detectability in FDG PET compared to the default, standard definition 4mm OSEM PET when using next generation digital PET/CT technology.