Motivation: FDG-PET, essential for identifying hypometabolism in epilepsy, is costly and involves radioactive tracers. To address this, we developed and validated a conditional GAN pipeline to produce FDG-PET from functional MRI using multi-device, multi-modal datasets. Goal(s): Develop a GAN-based MRI-to-PET translation framework and validate its clinical potential. Approach: Develop a deep learning model based on T1-weighted and blood oxygen level-dependent (BOLD) imaging to generate FDG-PET. Evaluate and validate the model using features including visual inspection, computational vision, standardized uptake value ratio, asymmetry index, and radiomics. Results: The synthetic PET scans demonstrate potential in enhancing epilepsy detection and predictive clinical outcomes. Impact: Deep learning can generate high-fidelity synthetic PET based on functional MRI Radiologists confirm high imaging fidelity between synthetic and actual PET The imaging features of synthetic PET are similar to actual PET Synthetic PET improves epilepsy diagnosis and prognosis
Yao et al. (Tue,) studied this question.