Estimating bone marrow adiposity from head MRI and identifying its genetic architecture.

A neural network can localize calvarial bone marrow adiposity from T1-weighted head MRI, enabling detection of 41 SNP-associated genetic loci, including six sex-specific loci.

• Why it matters: Bone marrow adiposity shifts across the human lifespan, but its characterization and genetic basis have been limited. Head MRI datasets offer an underused route to quantify this trait in large human samples.
• What they did: The study trained an artificial neural network to segment calvarial BM on T1-weighted head scans, then modeled sex-dimorphic age trajectories including menopause effects across several MRI datasets. It estimated heritability from SNP and twin data, performed genome-wide association to find 41 loci (six sex-specific), and integrated mapped genes with BM single-cell RNA-sequencing.
• The result: The trait showed high heritability and was linked to cognitive ability and Parkinson's disease via significant genetic correlations, supporting biology involving adipogenic differentiation and lipid loading.

AI summary of the abstract · may contain errors · not medical advice · read the paper.