Our deep learning based denoising solution for low-count PET has been rigorously validated across multiple tracers, in a multi-center, multi-reader setting. The results demonstrate strong generalizability and preserved diagnostic accuracy even at 50% count levels.

When comparing clinical scans with NUCLARITY-enhanced scans, we observed:

• Excellent quantitative agreement with SUVmean and SUVmax (CCC 1.00 and 0.99)
• Comparable diagnostic image quality and diagnostic confidence
• High diagnostic performance, with 99% sensitivity and 99% specificity for lesion detection
• Interscan agreement surpassing inter-reader variability

We want to explicitely thank our co-authors for their rigorous work in evaluating image quality, the reader study and for providing this strong manuscript: Justine Maes, Charles Carron, Simon DeKeyser, Tomas Brants, Vicky De Ridder, Amine Chaouki, Camille Steenhout, Stefaan Vandenberghe, Yves D'Asseler, Laurens Raes, Azzam Abdalla Ibrahim, gerard Moulin-Romsee, Isaac Kargar Samani, Ludovic D'hulst, Sezgin Üstmert and Maarten Larmuseau.

‍