A unified deep learning framework for cross-platform harmonization of multi-tracer PET quantification in neurodegenerative disease - PubMed
6 hours ago
- #Neurodegenerative Disease
- #Deep Learning
- #PET Harmonization
- Developed a unified deep learning framework to harmonize PET-MRI quantification to PET-CT standards for multiple tracers and scanner manufacturers.
- Reduced cross-platform bias by over 80% in paired same-day scans while preserving inter-regional biological topology.
- Generalized zero-shot to held-out tracers without retraining and validated across multicenter data, reducing discrepancies like amyloid Centiloid from 23.6 to 4.1.
- Aims to enable consistent cross-platform diagnostic cut-offs and reliable longitudinal monitoring when patients transition between modalities, supporting radiation-sparing quantitative PET in therapeutic workflows.