IP08-25 SUB-5-MINUTE DENOISED PET IMAGING OF INDETERMINATE RENAL MASSES USING ⁸⁹ Zr-DFO-Girentuximab: A FEASIBILITY STUDY

Introduction and Objectives: Incidental renal masses (< 7 cm) are increasingly detected on diagnostic images. While many small renal lesions are presumed to represent clear-cell renal cell carcinomas (ccRCC), up to 30% are benign at biopsy or surgery. Positron emission tomography (PET) imaging with 89Zr-DFO-girentuximab shows high sensitivity and specificity for ccRCC (>85%) [1]. However, the low administered dose (37 MBq) combined with delayed imaging (3-7 days post-injection, p.i.) required for clearance contributes to increased image noise and long scan times (~20 mins per bed position) on standard PET scanners. Advances in image denoising have shown potential to maintain diagnostic quality from noisy data in other clinical applications, supporting similar benefits for faster 89Zr-PET imaging in renal mass evaluation. Methods: Five patients with a single renal mass (< 7 cm) were injected with 37 MBq (±10%) of 89Zr-DFO girentuximab. Patients were scanned 5 ±2 days p.i. for 10 min on the PennPET Explorer (142 cm axial fieldof- view, time-of-flight PET/CT). The 10-min list-mode data were sub-sampled and reconstructed into 5- and 1-min datasets. Each image was post-processed by a deep learning denoising (DL-DN) model [2], trained on 6,500 18F-FDG image slices. An additional image set was obtained by 5-mm Gaussian-filtering of the subsampled images. Signal-to-noise (SNR; mean lesion counts relative to background noise) and contrast-to-noise (CNR; lesion-to-background count difference relative to background noise) ratios were measured using the contralateral non-lesional kidney as background. A reader study is ongoing to assess lesion presence, confidence rating, and image quality. Results: Both DL-DN and Gaussian-filtering reduce noise in the 5- and 1-min images. Gaussian-filtering noticeably blurs lesion details, while DL-DN improves sharpness and lesion conspicuity. DL-DN achieved the highest SNR and CNR for both reduced durations, while Gaussian-filtering showed only moderate improvement over the non-denoised data. For 5-mins images, the DL-DN SNR and CNR values exceeded those of the 10-min reference (dashed line). Conclusions: These results support DL-DN as a feasible method for reducing scan times in 89Zr-PET imaging, enabling sub-5-min scans without diagnostic image quality loss for renal mass evaluation. References: [1] Shuch B, Pantuck AJ, Bernhard JC, et al. (2024) Lancet Oncol. 25(10):1277-87 [2] Muller FM, Li EJ, Witherspoon ME, et al. (2025) EJNMMI. 12:1-9.