Diffusion Bridge Networks
Simulate Clinical-grade PET from MRI
for Dementia Diagnostics

Yitong Li1,2, Ralph Buchert3, Benita Schmitz-Koep4, Timo Grimmer5, Björn Ommer2,6, Dennis M. Hedderich‡4, Igor Yakushev‡7 Christian Wachinger‡1,2
‡Shared senior authors.
1Lab for AI in Medical Imaging, Technical University of Munich (TUM), Germany 2Munich Center for Machine Learning (MCML), Germany 3Department of Nuclear Medicine, University Medical Center Hamburg-Eppendorf, Germany 4Department of Neuroradiology, TUM University Hospital, Germany 5Department of Neurology, TUM University Hospital, Germany 6CompVis, Ludwig-Maximilians-University Munich (LMU), Germany 7Department of Nuclear Medicine, TUM University Hospital, Germany
Paper arXiv Code

SiM2P simulates PET from MRI with high fidelity

(AD: Alzheimer's Disease, FTD: Frontotemporal Dementia, CN: Normal Control)

82 years old female with AD

66 years old female with AD

71 years old male with FTD

60 years old female with FTD

69 years old female as CN

70 years old male as CN

Diffusion Bridge Framework


SiM2P employs a 3D diffusion bridge to simulate PET (SimPET) from structural MRI, conditioned on available auxiliary data. We validated the diagnostic utility of SimPET in a blinded clinical reader study, where SimPET showed significantly higher accuracy than MRI.

Enhanced Diagnostic Accuracy

Improved Diagnostic Accuracy


In a blinded clinical reader study, SimPET from SiM2P improved diagnostic accuracy by up to 12% over MRI across a range of tasks, with substantially increased interrater reliability.

Enhanced Diagnostic Accuracy

High Pathological Sensitivity


SiM2P offers high pathological sensitivity even in scenarios where structural MRI lacks sufficient pathological atrophy, such as in early-stage or atypical presentations of AD.

SiM2P with high pathological specificity

Local Deployment


With Local-Adapt, clinical sites can calibrate publicly pre-trained SiM2P models with as few as 20 local samples, while maintaining high-quality and accurate site-specific PET synthesis.

local adapt performance

Abstract

Positron emission tomography (PET) with 18-Fluorodeoxyglucose (FDG) is an established tool in the diagnostic workup of patients with suspected dementing disorders. However, compared to the routinely available magnetic resonance imaging (MRI), FDG-PET remains significantly less accessible and substantially more expensive.

Here, we present SiM2P, a 3D diffusion bridge-based framework that learns a probabilistic mapping from MRI and auxiliary patient information to simulate FDG-PET images of diagnostic quality.

In a blinded clinical reader study, two neuroradiologists and two nuclear medicine physicians rated the original MRI and SiM2P-simulated PET images of patients with Alzheimer's disease, behavioral-variant frontotemporal dementia, and cognitively healthy controls. SiM2P significantly improved the overall diagnostic accuracy of differentiating between three groups from 75.0% to 84.7% (p<0.05). Notably, the simulated PET images received higher diagnostic certainty ratings and achieved superior interrater agreement compared to original MRI images alone.

Finally, we developed a practical workflow for local deployment of the SiM2P framework. It requires as few as 20 site-specific cases and only basic demographic information. This approach makes the established diagnostic benefits of FDG-PET imaging more accessible to patients with suspected dementing disorders, potentially improving early detection and differential diagnosis in resource-limited settings.

BibTeX


      @article{li2025diffusion,
              title={Diffusion Bridge Networks Simulate Clinical-grade PET from MRI for Dementia Diagnostics},
              author={Li, Yitong and Buchert, Ralph and Schmitz-Koep, Benita and Grimmer, Timo and Ommer, Bj{\"o}rn and Hedderich, Dennis M and Yakushev, Igor and Wachinger, Christian},
              journal={arXiv preprint arXiv:2510.15556},
              year={2025}
              }