Objective:To construct a prior based on the inherent properties of PET to accurately segment the lesion areas.Methods:A network framework for PET tumor segmentation driven by geodesic priors was proposed (geodesic network for short). Specifically, partial differential equations were constructed to characterize the geodesic distances between different regions in PET images. Tumor marker points identified by CT labeling were used as the initial conditions for the equations. To enhance the contrast between areas of lung or breast tumors and normal tissues, a smooth Heaviside function was utilized to map the geodesic distances. The network framework adopted a dual-branch architecture, using geodesic priors to assist in PET image segmentation.Results:The proposed method achieved a Dice coefficient of 94.92% in lung cancer segmentation and 90.12% in breast cancer segmentation. With the addition of geodesic priors in the Unet, the Dice coefficient for breast cancer increased by 32.37% (from 42.50% to 74.87%).Conclusion:Geodesic priors can significantly improve segmentation outcomes and enhance the generalization capability of the network.

Objective:To assess the effectiveness of kinetic parameters of 18F-prostate specific membrane antigen (PSMA)-1007 PET/MR in distinguishing tumor recurrence (TR) from radiation necrosis (RN) in glioma patients. Methods:From January 2023 to June 2023, imaging data of 10 patients (6 males, 4 females; media age of 39.5 years) with gliomas who were suspected of recurrence and were referred for 18F-PSMA-1007 PET/MR scans at Sun Yat-Sen University Cancer Center were retrospectively analyzed. Static parameters from 18F-PSMA-1007 PET scans, including SUV max, SUV mean, metabolic tumor volume (MTV), and total lesion′s PSMA (TLP), as well as dynamic parameters including K 1, k 2, k 3, k 4, net influx rate (K i), and volume of distribution ( Vt) were obtained by using compartmental and multigraphical models. Additionally, parameters from dynamic contrast-enhanced MRI (DCE-MRI) were collected. Mann-Whitney U test was used to compare parameter differences between TR and RN groups. Spearman rank correlation analysis was used to explore the correlation between K i and DCE-MRI parameters. Results:Finally, 8 cases were diagnosed as TR and 2 cases were diagnosed as RN. The kinetic compartmental model-based evaluation determined that irreversible 2-tissue model (2T3K) provided the best-fitting results. The differences in SUV mean (median: 2.48 vs 0.89; Z=-2.09, P=0.044), SUV max (median: 4.04 vs 1.40; Z=-2.09, P=0.044), and K i (median: 1.33×10 -2vs 3.87×10 -3;Z=-2.10, P=0.044) between TR and RN groups were statistically significant. Some parameters of DCE-MRI also showed differences between the two groups ( Z=-2.09, P=0.044 for all). The K i yielded moderate correlation with DCE-MRI parameter Ve ( rs=0.650, P=0.042), while correlations between K i and other DCE-MRI parameters were not significant ( rs values: from -0.207 to 0.632, all P>0.05). Conclusion:Dynamic and multi-parametric 18F-PSMA-1007 PET/MR system holds promise for accurately distinguishing TR from RN in treated glioma patients.

Objective:To investigate the denoising performance of different deep learning (DL) strategies on low-dose brain 18F-FDG PET images. Methods:This retrospective methodological study was conducted on brain PET/CT images of 50 patients (35 males, 15 females, age 20-87 years) who received 3.7MBq/kg 18F-FDG at the Fifth Affiliated Hospital of Sun Yat-sen University between May 2023 and January 2024. Full-dose PET data were acquired with 2min scan. CT scans were acquired before PET scanning. Low-dose PET sinograms were generated by down-sampling the full-dose list mode data to 1/2, 1/4, and 1/20 of full-dose count level. Both full-dose and low-dose sinograms were reconstructed with random, CT-based attenuation and scatter corrections using the three-dimensional (3D) ordered-subsets expectation maximization (OSEM) algorithm (2 iterations, 20 subsets). A total of 4 DL denoising methods were established: (1) 3D conditional generative adversarial networks (GAN) using only low-dose PET as input (GAN-1); (2) 3D attention-based GAN (AttGAN) with low-dose PET input (AttGAN-1); (3) 3D AttGAN with low-dose PET and CT inputs (AttGAN-2); (4) 3D AttGAN with frequency-separation using low-dose PET and CT inputs (AttGAN-FS-2). For AttGAN-FS-2, during the frequency division process, high- and low-frequency components were extracted from the PET reconstructed images via Fourier transform, then inversed Fourier transform, denoised separately, and finally combined to produce the final denoised images. The dataset was separated into training (70%), validation (10%) and testing (20%) sets using simple random sampling without replacement with a fixed random seed. A 5-fold cross-validation scheme was then applied to test all 50 patients. Performance was evaluated against full-dose PET using normalized mean square error (NMSE), structural similarity (SSIM), peak signal-to-noise ratio (PSNR), contrast-to-noise ratio (CNR), SUV mean and SUV max bias of selected brain ROIs. Wilcoxon signed rank test was used to analyze the differences between the denoising methods. Results:AttGAN-FS-2 showed the best performance among all dose levels, with statistical difference as compared by low-dose PET and GAN-1 denoised images for NMSE, SSIM, PSNR, and CNR ( Z values: 2.92-6.15, all P<0.005). NMSE, SSIM quantitative evaluation results (median) of each model at 1/20 dose were: GAN-1: 0.08, 0.87, AttGAN-1: 0.08, 0.88, AttGAN-2: 0.07, 0.89, AttGAN-FS-2: 0.06, 0.91, respectively ( Z values: 3.24-5.77, all P<0.005). Conclusion:The DL-based method combined with multiple strategies AttGAN-FS-2 shows improved denoising performance for low-dose brain PET images.

Objective:To construct a prior based on the inherent properties of PET to accurately segment the lesion areas.Methods:A network framework for PET tumor segmentation driven by geodesic priors was proposed (geodesic network for short). Specifically, partial differential equations were constructed to characterize the geodesic distances between different regions in PET images. Tumor marker points identified by CT labeling were used as the initial conditions for the equations. To enhance the contrast between areas of lung or breast tumors and normal tissues, a smooth Heaviside function was utilized to map the geodesic distances. The network framework adopted a dual-branch architecture, using geodesic priors to assist in PET image segmentation.Results:The proposed method achieved a Dice coefficient of 94.92% in lung cancer segmentation and 90.12% in breast cancer segmentation. With the addition of geodesic priors in the Unet, the Dice coefficient for breast cancer increased by 32.37% (from 42.50% to 74.87%).Conclusion:Geodesic priors can significantly improve segmentation outcomes and enhance the generalization capability of the network.

Objective:To assess the effectiveness of kinetic parameters of 18F-prostate specific membrane antigen (PSMA)-1007 PET/MR in distinguishing tumor recurrence (TR) from radiation necrosis (RN) in glioma patients. Methods:From January 2023 to June 2023, imaging data of 10 patients (6 males, 4 females; media age of 39.5 years) with gliomas who were suspected of recurrence and were referred for 18F-PSMA-1007 PET/MR scans at Sun Yat-Sen University Cancer Center were retrospectively analyzed. Static parameters from 18F-PSMA-1007 PET scans, including SUV max, SUV mean, metabolic tumor volume (MTV), and total lesion′s PSMA (TLP), as well as dynamic parameters including K 1, k 2, k 3, k 4, net influx rate (K i), and volume of distribution ( Vt) were obtained by using compartmental and multigraphical models. Additionally, parameters from dynamic contrast-enhanced MRI (DCE-MRI) were collected. Mann-Whitney U test was used to compare parameter differences between TR and RN groups. Spearman rank correlation analysis was used to explore the correlation between K i and DCE-MRI parameters. Results:Finally, 8 cases were diagnosed as TR and 2 cases were diagnosed as RN. The kinetic compartmental model-based evaluation determined that irreversible 2-tissue model (2T3K) provided the best-fitting results. The differences in SUV mean (median: 2.48 vs 0.89; Z=-2.09, P=0.044), SUV max (median: 4.04 vs 1.40; Z=-2.09, P=0.044), and K i (median: 1.33×10 -2vs 3.87×10 -3;Z=-2.10, P=0.044) between TR and RN groups were statistically significant. Some parameters of DCE-MRI also showed differences between the two groups ( Z=-2.09, P=0.044 for all). The K i yielded moderate correlation with DCE-MRI parameter Ve ( rs=0.650, P=0.042), while correlations between K i and other DCE-MRI parameters were not significant ( rs values: from -0.207 to 0.632, all P>0.05). Conclusion:Dynamic and multi-parametric 18F-PSMA-1007 PET/MR system holds promise for accurately distinguishing TR from RN in treated glioma patients.

Objective:To investigate the denoising performance of different deep learning (DL) strategies on low-dose brain 18F-FDG PET images. Methods:This retrospective methodological study was conducted on brain PET/CT images of 50 patients (35 males, 15 females, age 20-87 years) who received 3.7MBq/kg 18F-FDG at the Fifth Affiliated Hospital of Sun Yat-sen University between May 2023 and January 2024. Full-dose PET data were acquired with 2min scan. CT scans were acquired before PET scanning. Low-dose PET sinograms were generated by down-sampling the full-dose list mode data to 1/2, 1/4, and 1/20 of full-dose count level. Both full-dose and low-dose sinograms were reconstructed with random, CT-based attenuation and scatter corrections using the three-dimensional (3D) ordered-subsets expectation maximization (OSEM) algorithm (2 iterations, 20 subsets). A total of 4 DL denoising methods were established: (1) 3D conditional generative adversarial networks (GAN) using only low-dose PET as input (GAN-1); (2) 3D attention-based GAN (AttGAN) with low-dose PET input (AttGAN-1); (3) 3D AttGAN with low-dose PET and CT inputs (AttGAN-2); (4) 3D AttGAN with frequency-separation using low-dose PET and CT inputs (AttGAN-FS-2). For AttGAN-FS-2, during the frequency division process, high- and low-frequency components were extracted from the PET reconstructed images via Fourier transform, then inversed Fourier transform, denoised separately, and finally combined to produce the final denoised images. The dataset was separated into training (70%), validation (10%) and testing (20%) sets using simple random sampling without replacement with a fixed random seed. A 5-fold cross-validation scheme was then applied to test all 50 patients. Performance was evaluated against full-dose PET using normalized mean square error (NMSE), structural similarity (SSIM), peak signal-to-noise ratio (PSNR), contrast-to-noise ratio (CNR), SUV mean and SUV max bias of selected brain ROIs. Wilcoxon signed rank test was used to analyze the differences between the denoising methods. Results:AttGAN-FS-2 showed the best performance among all dose levels, with statistical difference as compared by low-dose PET and GAN-1 denoised images for NMSE, SSIM, PSNR, and CNR ( Z values: 2.92-6.15, all P<0.005). NMSE, SSIM quantitative evaluation results (median) of each model at 1/20 dose were: GAN-1: 0.08, 0.87, AttGAN-1: 0.08, 0.88, AttGAN-2: 0.07, 0.89, AttGAN-FS-2: 0.06, 0.91, respectively ( Z values: 3.24-5.77, all P<0.005). Conclusion:The DL-based method combined with multiple strategies AttGAN-FS-2 shows improved denoising performance for low-dose brain PET images.