Related Experiment Video
Updated: May 12, 2025
05:17
Hybrid PET/MRI Imaging of Alzheimer's Disease Based on 18F-AV-1451
Published on: April 18, 2025
65
MRI2PET: Realistic PET Image Synthesis from MRI for Automated Inference of Brain Atrophy and Alzheimer's
Brandon Theodorou1,2, Anant Dadu2,3, Brian Avants4
1Department of Computer Science, University of Illinois at Urbana-Champaign, Urbana, IL, USA.
Summary
This study introduces MRI2PET, a novel method using generative AI to create Positron Emission Tomography (PET) scans from Magnetic Resonance Imaging (MRI). This approach enhances disease classification accuracy, making advanced brain imaging more accessible.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neuroscience
Background:
- Positron Emission Tomography (PET) scans are vital for diagnosing complex conditions like cancer and cognitive disorders.
- PET scans are significantly more expensive than other imaging modalities like MRI, limiting their accessibility.
- High-quality PET imaging is crucial for medical applications and machine learning research.
Purpose of the Study:
- To develop a cost-effective method for generating PET scans from MRI data.
- To address the challenges of limited paired data and complex 3D image nuances in PET synthesis.
- To improve disease classification accuracy using AI-generated PET scans.
Main Methods:
- Proposed MRI2PET, a 3D diffusion-based generative model.
- Utilized style-transferred pre-training and a Laplacian pyramid loss.
- Leveraged larger unpaired MRI datasets and structural similarities between MRI and PET images.
Main Results:
- MRI2PET successfully generated realistic AV45-PET scans from T1-weighted MRI.
- Augmentation with MRI2PET improved the Area Under the Receiver Operating Characteristic Curve (AUROC) for brain scan classification from 0.688 to 0.780.
- The model demonstrated improved classification across cognitively normal, mild cognitive impairment, and Alzheimer's Disease groups.
Conclusions:
- The ability to generate high-quality PET scans from MRI can expand accessible imaging workflows.
- This technology has the potential to enhance machine learning capabilities in medical imaging.
- Improved PET scan generation can lead to better patient care and diagnostics.
