Recent genome wide association studies have revealed that several osteoarthritis (OA) risk loci involve common genetic variations related to musculoskeletal development and morphogenesis. To date, a major shortcoming, is that morphometric phenotyping is based on 2D superposition imaging, resulting in high noise levels and limited applicability. In this interdisciplinary project, DNA will be collected from patients undergoing CT scanning for medical purposes. Advanced image processing and deep learning methods will be employed for 3D phenotyping and to establish the association between SNP genotype and the joint shape and mechanical phenotype at a population wide level.

The proposed methodology builds on most recent advances in geometric deep learning and shape modeling to enable reliable computation of joint contact stresses. Findings will be benchmarked by identification of high-load bearing zones in patients having early-stage OA using a large validation cohort of +1200 OA cases.