Clinical Scorecard: Deep Learning-Based Synthesis of MRI and PET for Quantifying Amyloid-β in Alzheimer’s Disease

At a Glance

Key Highlights

• Aβ PET imaging strongly correlates with neuropathology and aids early diagnosis and disease monitoring.
• Structural MRI combined with deep learning (GAN) can synthesize Aβ PET images non-invasively.
• GAN-based MRI-to-PET synthesis offers a cost-effective, accessible alternative to direct PET imaging.

Guideline-Based Recommendations

Diagnosis

• Use CSF Aβ42/40 biomarkers for high diagnostic accuracy but consider invasiveness.
• Plasma p-tau217/Aβ42 ratio shows promise but requires confirmation with PET due to standardization limitations.
• Aβ PET imaging is recommended for in vivo visualization and quantification of cerebral Aβ pathology.

Management

• Quantitative Aβ PET indices in specific brain regions are key for monitoring treatment efficacy.
• Structural MRI is routinely used to assess cortical atrophy associated with Aβ deposition.

Monitoring & Follow-up

• Use quantitative Aβ PET to track disease progression and response to disease-modifying therapies.
• GAN-synthesized Aβ PET images from MRI may support identification of patients needing confirmatory PET scans.

Risks

• PET imaging involves radiation exposure and high costs limiting accessibility.
• CSF biomarker analysis requires invasive lumbar puncture.
• Plasma biomarkers currently lack cross-platform standardization for standalone diagnosis.

Patient & Prescribing Data

Individuals across the AD spectrum including healthy controls, MCI, AD, and other dementias

GAN-based MRI-to-PET synthesis may reduce need for invasive or costly procedures, aiding early diagnosis and monitoring

Clinical Best Practices

• Combine structural MRI with deep learning models to generate synthetic Aβ PET images for non-invasive assessment.
• Confirm plasma biomarker findings with PET imaging due to current standardization limitations.
• Utilize quantitative PET metrics for evaluating disease progression and therapeutic efficacy.
• Apply standardized preprocessing protocols for MRI and PET data to ensure accurate image registration and analysis.

References

• Alzheimer’s disease pathology and biomarkers overview
• CSF and plasma biomarkers in AD diagnosis
• Aβ PET imaging correlation with neuropathology
• GAN models for MRI to PET image synthesis
• ADNI database and imaging protocols