3D generative model synthesizes FDG-PET from T1 MRI and ASL in Alzheimer's disease

A 3D generative AI model synthesized FDG-PET from T1-weighted MRI and arterial spin labeling in Alzheimer's disease, achieving >93% structural similarity to real PET and generalizing to an independent external cohort.

• Training used paired MRI and FDG-PET acquisitions: T1-weighted MRI alone in 1170 cases and T1-weighted MRI with ASL in 220 cases.
• Adding arterial spin labeling perfusion to T1-weighted MRI provided modest improvements in AD-specific regions.
• The model's diagnostic performance was evaluated with blinded reader studies and differential diagnosis in independent external cohorts.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.