Deep learning-based attenuation correction for whole-body PET. A multi-tracer study with ${}^{18}$F-FDG, ${}^{68}$Ga-DOTATATE, and ${}^{18}F$-Fluciclovine

A novel deep learning (DL)-based attenuation correction (AC) framework was applied to clinical whole-body oncology studies using ${}^{18}$F-FDG, ${}^{68}$Ga-DOTATATE, and ${}^{18}$F-Fluciclovine. The framework used activity (λ-MLAA) and attenuation (µ-MLAA) maps estimated by the maximum likelihood reconstruction of activity and attenuation (MLAA) algorithm as inputs to a modified U-net neural network with a novel imaging physics-based loss function to learn a CT-derived attenuation map (µ-CT). Clinical whole-body PET/CT datasets of ${}^{18}$F-FDG (N = 113), ${}^{68}$Ga-DOTATATE (N = 76), and ${}^{18}$F-Fluciclovine (N = 90) were used to train and test tracer-specific neural networks. For each tracer, forty subjects were used to train the neural network to predict attenuation maps (µ-DL). µ-DL and µ-MLAA were compared to the gold-standard µ-CT. PET images reconstructed using the OSEM algorithm with µ-DL (OSEM${}_{DL}$) and µ-MLAA (OSEM${}_{MLAA}$) were compared to the CT-based reconstruction (OSEM${}_{CT}$). Tumor regions of interest were segmented by two radiologists and tumor SUV and volume measures were reported, as well as evaluation using conventional image analysis metrics. µ-DL yielded high resolution and fine detail recovery of the attenuation map, which was superior in quality as compared to µ-MLAA in all metrics for all tracers. Using OSEM${}_{CT}$ as the gold-standard, OSEM${}_{DL}$ provided more accurate tumor quantification than OSEM${}_{MLAA}$ for all three tracers, e.g., error in SUV${}_{max}$ for OSEM${}_{MLAA}$ vs. OSEM${}_{DL}$: - 3.6 ± 4.4% vs. - 1.7 ± 4.5% for ${}^{18}$F-FDG (N = 152), - 4.3 ± 5.1% vs. 0.4 ± 2.8% for ${}^{68}$Ga-DOTATATE (N = 70), and - 7.3 ± 2.9% vs. - 2.8 ± 2.3% for ${}^{18}$F-Fluciclovine (N = 44). OSEM${}_{DL}$ also yielded more accurate tumor volume measures than OSEM${}_{MLAA}$, i.e., - 8.4 ± 14.5% (OSEM${}_{MLAA}$) vs. - 3.0 ± 15.0% for ${}^{18}$F-FDG, - 14.1 ± 19.7% vs. 1.8 ± 11.6% for ${}^{68}$Ga-DOTATATE, and - 15.9 ± 9.1% vs. - 6.4 ± 6.4% for ${}^{18}$F-Fluciclovine. The proposed framework provides accurate and robust attenuation correction for whole-body ${}^{18}$F-FDG, ${}^{68}$Ga-DOTATATE and ${}^{18}$F-Fluciclovine in tumor SUV measures as well as tumor volume estimation. The proposed method provides clinically equivalent quality as compared to CT in attenuation correction for the three tracers.