Synthesizing Multi-tracer PET Images for Alzheimer’s Disease Patients Using a 3D Unified Anatomy-Aware Cyclic Adversarial Network

Positron Emission Tomography (PET) is an important tool for studying Alzheimer’s disease (AD). PET scans can be used as diagnostics tools, and to provide molecular characterization of patients with cognitive disorders. However, multiple tracers are needed measure glucose metabolism (\(^{18}\)F-FDG), synaptic vesicle protein (\(^{11}\)C-UCB-J), \(\beta \)-amyloid (\(^{11}\)C-PiB). Administering patient will lead high radiation dose cost. In addition, access using new or less-available sophisticated production methods short half-life isotopes may very limited. Thus, it desirable develop efficient multi-tracer synthesis model that generate from single-tracer PET. Previous works on medical image focus one-to-one fixed domain translations, cannot simultaneously learn the feature domains. Given 3 more tracers, relying previous also create heavy burden number models trained. To tackle these issues, we propose a 3D unified anatomy-aware cyclic adversarial network (UCAN) translating volumes one generative model, where MR anatomical information incorporated. Evaluations dataset demonstrate feasibility our UCAN high-quality volumes, NMSE less than \(15\%\) all tracers. Our code available at https://github.com/bbbbbbzhou/UCAN.