Generation of Heterogeneous PET Images from Uniform Organ Activity Maps Using a Pretrained Domain-Adapted Diffusion Model
arXiv:2605.20267v1 Announce Type: cross Abstract: Synthetic PET images are valuable for quantitative imaging workflow development, scalable virtual imaging trials, and deep learning model training, but conventional physics-based simulation approaches are computationally intensive, limited in anatomical variability, and often fail to capture heterogeneous PET uptake. This study developed a pretrained domain-adapted diffusion (PAD) model for anatomy-conditioned PET synthesis from uniform organ activity maps. PAD adopts a natural-image pretrained text-to-image decoder with an upstream conditionin
Advances in diffusion models and domain adaptation techniques have matured, allowing for accurate and efficient synthetic medical image generation.
This development can significantly accelerate medical imaging research, reduce costs associated with physical trials, and improve the training data for AI models in healthcare.
The ability to generate high-fidelity synthetic PET images from basic organ activity maps changes the paradigm for medical image simulation and AI model development in diagnostics.
- · Medical AI developers
- · Pharmaceutical companies (drug discovery)
- · Hospitals and research institutions (training)
- · Companies reliant on expensive physical medical imaging trials
Reduced computational time and cost for generating medical imaging data for research and development.
Faster development and deployment of new diagnostic AI models due to abundant and diverse synthetic training data.
Potential for personalized medicine advancements through highly customized synthetic patient data for treatment planning and prediction.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI