MRAβ: A multimodal MRI-derived amyloid-β biomarker for Alzheimer's disease

Area of Science:

• Neuroimaging
• Biomarker Development
• Artificial Intelligence in Medicine

Background:

• Florbetapir ¹⁸ F (AV45) PET imaging is a sensitive biomarker for Alzheimer's disease (AD) amyloid-β but is limited by radiation exposure and cost.
• Developing non-irradiating, cost-effective biomarkers is crucial for widespread AD diagnosis and monitoring.

Purpose of the Study:

• To develop a multimodal MRI-derived Amyloid-β (MRAβ) biomarker using a collaborative generative adversarial networks (CollaGAN) model.
• To evaluate MRAβ's ability to mimic AV45 PET imaging and its diagnostic performance in distinguishing Alzheimer's disease and mild cognitive impairment from healthy controls.

Main Methods:

• Collected multimodal MRI and AV45 PET data from 380 participants in the ADNI dataset and 64 in the OASIS3 dataset.
• Applied a five-fold cross-validation CollaGAN model to generate MRAβ from multimodal MRI data.
• Utilized voxel-wise t-tests and a 3D ResNet classifier to assess MRAβ's accuracy and comparability to AV45.

Main Results:

• MRAβ successfully characterized subject-level AV45 spatial variations in AD and MCI patients.
• Amyloid-β depositions identified by MRAβ in AD and MCI were significantly higher than in healthy controls and highly correlated with AV45 findings (r > .92).
• MRAβ demonstrated comparable performance to AV45 in discriminating AD from healthy controls and MCI from healthy controls.

Conclusions:

• Multimodal MRI-derived MRAβ can effectively mimic cerebral amyloid-β depositions detected by AV45 PET imaging.
• MRAβ shows promise as a radiation-free, cost-effective molecular-explainable biomarker for Alzheimer's disease.
• This approach supports the advancement of MRI techniques for molecular biomarker applications in neurodegenerative diseases.