Brain-age in ultra-low-field MRI: how well does it work?

Area of Science:

• Neuroimaging
• Biomarkers
• Magnetic Resonance Imaging (MRI)

Background:

• Brain-age estimation, derived from neuroimaging, serves as a biomarker for brain health and disease risk.
• High-field (HF) MRI is typically used but is costly and inaccessible.
• Ultra-low-field (ULF) MRI offers a cheaper, more accessible alternative, but its lower resolution poses challenges for biomarker reliability.

Purpose of the Study:

• To systematically evaluate different brain-age estimation pipelines using ULF MRI.
• To compare the performance of ULF MRI brain-age estimation against HF MRI.
• To assess the validity, correspondence, and reliability of ULF brain-age estimates.

Main Methods:

• 23 adults were scanned using one HF MRI and two ULF MRI systems.
• 14 distinct acquisition types (varying T1/T2-weighting, resolution, preprocessing) were analyzed.
• Five brain-age software packages were employed to process 573 scans, evaluating validity, HF correspondence, and test-retest reliability.

Main Results:

• Several ULF MRI pipelines achieved performance comparable to HF MRI.
• The best-performing pipelines demonstrated moderate-to-strong validity (r=0.76-0.92) and correspondence to HF (r=0.84-0.93).
• Excellent test-retest reliability (r=0.97-0.99) was observed across multiple ULF pipelines, with some anisotropic acquisitions performing comparably to enhanced images.

Conclusions:

• Accurate and reliable brain-age estimates are achievable with ULF MRI across various pipelines, often without requiring image enhancement.
• Pipeline performance is dependent on the specific combination of model, scan type, and preprocessing.
• ULF brain-age estimation presents a practical, scalable tool for accessible neuroimaging biomarkers in clinical settings and research.