Artificial intelligence-based PET denoising could allow a two-fold reduction in [18F]FDG PET acquisition time in digital PET/CT

Kathleen Weyts (1) , Charline Lasnon (1) , Renaud Ciappuccini (1) , Justine Lequesne (1) , Aurélien Corroyer-Dulmont (1, 2) , Elske Quak (1) , Bénédicte Clarisse (1) , Laurent Roussel (1) , Stéphane Bardet (1) , Cyril Jaudet (1)

1 UNICANCER/CRLC - Centre Régional de Lutte contre le Cancer François Baclesse [Caen]
2 ISTCT - Imagerie et Stratégies Thérapeutiques pour les Cancers et Tissus cérébraux

Purpose We investigated whether artificial intelligence (AI)-based denoising halves PET acquisition time in digital PET/CT. Methods One hundred ninety-five patients referred for [ 18 F]FDG PET/CT were prospectively included. Body PET acquisitions were performed in list mode. Original "PET90" (90 s/bed position) was compared to reconstructed ½-duration PET (45 s/bed position) with and without AI-denoising, "PET45AI and PET45". Denoising was performed by SubtlePET™ using deep convolutional neural networks. Visual global image quality (IQ) 3-point scores and lesion detectability were evaluated. Lesion maximal and peak standardized uptake values using lean body mass (SUL max and SUL peak), metabolic volumes (MV), and liver SUL mean were measured, including both standard and EARL 1 (European Association of Nuclear Medicine Research Ltd) compliant SUL. Lesion-to-liver SUL ratios (LLR) and liver coefficients of variation (CV liv) were calculated. Results PET45 showed mediocre IQ (scored poor in 8% and moderate in 68%) and lesion concordance rate with PET90 (88.7%). In PET45AI, IQ scores were similar to PET90 (P = 0.80), good in 92% and moderate in 8% for both. The lesion concordance rate between PET90 and PET45AI was 836/856 (97.7%), with 7 lesions (0.8%) only detected in PET90 and 13 (1.5%) exclusively in PET45AI. Lesion EARL 1 SUL peak was not significantly different between both PET (P = 0.09). Lesion standard SUL peak , standard and EARL1 SUL max , LLR and CV liv were lower in PET45AI than in PET90 (P < 0.0001), while lesion MV and liver SUL mean were higher (P < 0.0001). Good to excellent intraclass correlation coefficients (ICC) between PET90 and PET45AI were observed for lesion SUL and MV (ICC ≥ 0.97) and for liver SUL mean (ICC ≥ 0.87). Conclusion AI allows [ 18 F]FDG PET duration in digital PET/CT to be halved, while restoring degraded ½-duration PET image quality. Future multicentric studies, including other PET radiopharmaceuticals, are warranted.

Mots clés: 18 F]FDG; PET; Denoising; Artificial intelligence; Deep learning; Acquisition time

Domaines: Imagerie médicale; Intelligence artificielle [cs.AI]; Médecine nucléaire

Kathleen Weyts, Charline Lasnon, Renaud Ciappuccini, Justine Lequesne, Aurélien Corroyer-Dulmont, et al.. Artificial intelligence-based PET denoising could allow a two-fold reduction in [18F]FDG PET acquisition time in digital PET/CT. European Journal of Nuclear Medicine and Molecular Imaging, 2022, 49 (11), pp.3750-3760. ⟨10.1007/s00259-022-05800-1⟩. ⟨hal-03699159⟩