Cross-sectional and longitudinal Biomarker extraction and analysis for multicentre FLAIR brain MRI

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Radiology

Background:

• Fluid-attenuated inversion recovery (FLAIR) MRI is crucial for analyzing cerebrovascular disease (CVD) and Alzheimer's disease (AD).
• Automated extraction of cerebral biomarkers from FLAAR datasets is vital for disease characterization and monitoring.
• Current automated algorithms often require T1-weighted or multi-modal inputs, limiting FLAIR-only applications.

Purpose of the Study:

• To evaluate the feasibility of using FLAIR MRI-only datasets for automated biomarker extraction.
• To characterize healthy subjects, and those with AD and CVD using FLAIR-based biomarkers.
• To assess if FLAIR MRI can provide similar diagnostic information as traditional multi-modal approaches.

Main Methods:

• Utilized deep learning-based segmentation algorithms specifically designed for FLAIR MRI.
• Extracted cross-sectional biomarkers (total brain volume, CSF volume, white matter lesion volume) and longitudinal changes.
• Analyzed large datasets from dementia and vascular disease cohorts (over 200,000 images).

Main Results:

• Automated tools successfully extracted biomarkers from large, FLAIR-only datasets.
• FLAIR-derived biomarkers showed trends consistent with traditional modalities in differentiating healthy, AD, and CVD subjects.
• Demonstrated the potential for FLAIR MRI in end-to-end analysis of AD and CVD datasets.

Conclusions:

• FLAIR MRI is a viable sequence for end-to-end analysis of large Alzheimer's disease and cerebrovascular disease datasets.
• Using FLAIR-only data can lower acquisition costs, simplify clinical translation, and reduce measurement errors.
• This approach supports improved disease characterization and monitoring in large patient cohorts.