TY - GEN
T1 - Dual-Domain Equivariant Generative Adversarial Network for Multimodal CT-Pet Synthesis
AU - Steele, Gabriel
AU - Altalib, Alzahra
AU - Perelli, Alessandro
N1 - Publisher Copyright:
© 2026 IEEE.
PY - 2026/5/20
Y1 - 2026/5/20
N2 - We present a Dual-Domain Equivariant Generative Adversarial Network (DDE-GAN) for multimodal CT-PET image synthesis. Traditional GAN-based approaches often operate solely in the spatial domain and ignore geometric consistency, resulting in limited structural fidelity. DDE-GAN addresses these challenges by jointly learning from both spatial and frequency (Fourier) domains, capturing complementary anatomical and spectral information. Furthermore, rotational equivariance embedded in the physics of the CT and PET measurements are integrated into the loss of both the generator and discriminator to ensure consistent responses under rotations, improving anatomical accuracy. A hierarchical dual-domain training strategy enforces intraand inter-domain consistency through multi-stage loss functions. Evaluated on the HECKTOR 2022 CT-PET dataset, DDE-GAN achieves superior synthesis quality over baseline models for CT-PET image synthesis. The results demonstrate that combining dual-domain learning with geometric equivariance substantially enhances multimodal image synthesis accuracy and robustness, enabling practical applications in PET completion and data augmentation.
AB - We present a Dual-Domain Equivariant Generative Adversarial Network (DDE-GAN) for multimodal CT-PET image synthesis. Traditional GAN-based approaches often operate solely in the spatial domain and ignore geometric consistency, resulting in limited structural fidelity. DDE-GAN addresses these challenges by jointly learning from both spatial and frequency (Fourier) domains, capturing complementary anatomical and spectral information. Furthermore, rotational equivariance embedded in the physics of the CT and PET measurements are integrated into the loss of both the generator and discriminator to ensure consistent responses under rotations, improving anatomical accuracy. A hierarchical dual-domain training strategy enforces intraand inter-domain consistency through multi-stage loss functions. Evaluated on the HECKTOR 2022 CT-PET dataset, DDE-GAN achieves superior synthesis quality over baseline models for CT-PET image synthesis. The results demonstrate that combining dual-domain learning with geometric equivariance substantially enhances multimodal image synthesis accuracy and robustness, enabling practical applications in PET completion and data augmentation.
KW - CT-PET image Synthesis
KW - Dual-domain
KW - Equivariance
KW - Generative Adversarial Network
UR - https://www.scopus.com/pages/publications/105041634292
U2 - 10.1109/ISBI61048.2026.11515956
DO - 10.1109/ISBI61048.2026.11515956
M3 - Conference contribution
AN - SCOPUS:105041634292
T3 - Proceedings - International Symposium on Biomedical Imaging
BT - ISBI 2026 - 23rd IEEE International Symposium on Biomedical Imaging
PB - IEEE Computer Society
CY - London
T2 - 23rd IEEE International Symposium on Biomedical Imaging, ISBI 2026
Y2 - 8 April 2026 through 11 April 2026
ER -