CoCoLIT: ControlNet-Conditioned Latent Image Translation for MRI to Amyloid PET Synthesis

The introduction of CoCoLIT (ControlNet-Conditioned Latent Image Translation) marks a significant advancement in the synthesis of amyloid PET scans from MRI data, which is more widely available and accessible. This method is particularly important for large-scale Alzheimer's Disease screening, as it provides a cost-effective alternative for detecting amyloid pathology. CoCoLIT incorporates innovative techniques such as Weighted Image Space Loss and Latent Average Stabilization, which enhance the quality of latent representation learning and synthesis. Performance evaluations on publicly available datasets indicate that CoCoLIT significantly outperforms state-of-the-art methods, showcasing its potential to transform diagnostic practices in Alzheimer's care. By simplifying the translation of complex 3D neuroimaging data into actionable insights, CoCoLIT could facilitate earlier detection and intervention, ultimately improving patient outcomes in Alzheimer's Disease management.