MRI-to-PET synthesis via deep learning for amyloid-β quantification in Alzheimer’s disease

By
Zhigeng Chen
Sheng Bi
Yi Shan
Feng Wang
Yong Wang
Zhongyuan Qi
Tao Wang
Xiaoyuan Li
Shengnan Li
Huanhui Xiao
Silun Wang
Bixiao Cui
Zhigang Qi
Ying Han
Shaozhen Yan
Jie Lu
January 7, 2026
0 min

European Radiology

Objective:

To develop a GAN-based model for generating synthetic Aβ PET images from structural MRI and evaluate their similarity and quantitative consistency with real Aβ PET, highlighting its significance for early Alzheimer's diagnosis.

Key Findings:

The GAN model successfully generated synthetic Aβ PET images from structural MRI, with high similarity and quantitative consistency with real Aβ PET images, supported by statistical metrics.
The approach may serve as a non-invasive tool for identifying patients needing further PET confirmation.

Interpretation:

The study demonstrates the potential of using deep learning to synthesize Aβ PET images from MRI, which could enhance early diagnosis and monitoring of Alzheimer's disease, with significant clinical implications.

Limitations:

The study is retrospective and may have inherent biases, such as selection bias and confounding variables.
Further validation in diverse populations and clinical settings is necessary.

Conclusion:

The GAN-based synthesis of Aβ PET images from MRI presents a promising, non-invasive alternative for assessing amyloid-β pathology in Alzheimer's disease, warranting further research and clinical application to validate its efficacy.

MRI-to-PET synthesis via deep learning for amyloid-β quantification in Alzheimer’s disease

Objective:

Key Findings:

Interpretation:

Limitations:

Conclusion:

Original Source(s)

MRI-to-PET synthesis via deep learning for amyloid-β quantification in Alzheimer’s disease

Related Content

The Cortical Asymmetry Index for subtyping dementia patients

Redefining Neuro-Ophthalmology

Biomarker evidence of neurodegeneration in mid-life former rugby players