Biomarkers.

Area of Science:

• Neuroimaging
• Artificial Intelligence in Medicine
• Biomarker Discovery

Background:

• Traditional Alzheimer's disease (AD) pathology detection methods are costly, invasive, and inaccessible.
• Early identification of AD biomarkers is crucial for timely interventions and improved patient outcomes.
• A novel, cost-effective, non-invasive method using multi-modal MRI was developed to detect amyloid pathology, a key AD hallmark.

Purpose of the Study:

• To develop and validate a deep learning model for detecting amyloid pathology using MRI scans.
• To provide a safer, more affordable, and accessible alternative to existing AD biomarker detection methods.
• To leverage routinely acquired MRI scans for enhanced clinical insights in AD diagnosis.

Main Methods:

• A deep learning model was trained on 1,488 paired amyloid PET and MRI scans from the Alzheimer's Disease Neuroimaging Initiative (ADNI).
• T1-weighted and T2-weighted MRI scans served as input to synthesize images representative of amyloid pathology.
• Model performance was evaluated using sensitivity, specificity, and accuracy metrics on a validation dataset, comparing synthesized image SUVR scores to ground truth PET images.

Main Results:

• The deep learning model achieved a sensitivity of 0.70, specificity of 0.72, and accuracy of 0.71 in detecting amyloid positivity.
• Synthesized images showed strong alignment with original PET images, indicated by comparable SUVR scores.
• The model demonstrated clinically relevant accuracy in identifying amyloid positivity.

Conclusions:

• This study presents a proof-of-concept for a safer, more affordable, and accessible method for amyloid detection.
• The developed technology utilizes standard MRI scans to provide non-invasive insights for identifying candidates for early disease-modifying therapies.
• Further model refinement is ongoing, but the approach holds significant potential for improving early AD detection and patient prognosis.