Mass Univariate Regression Analysis for Three-Dimensional Liver Image-Derived Phenotypes

Image-derived phenotypes of abdominal organs from magnetic resonance imaging reveal variations in volume and shape may be used to model changes a normal versus pathological organ improve diagnosis. Computational atlases anatomical provide many advantages quantifying modeling differences size for population studies. Here we made use liver segmentations derived Dixon MRI 2,730 UK Biobank participants create 3D meshes. We computed the signed distances between reference subject-specific meshes define surface-to-surface (S2S) phenotype. employed mass univariate regression analysis compare S2S values image-derived specific liver, such as proton density fat fraction iron concentration while adjusting age, sex, ethnicity, body index waist-to-hip ratio. Vertex-based associations mesh were extracted threshold-free cluster enhancement was applied sensitivity stability statistical parametric maps. Our findings show that are robust method association anatomical, anthropometric, phenotypic across liver. This approach readily different clinical conditions well extended other larger population.