Automatic hip abductor muscle fat fraction estimation and association with early OA cartilage degeneration biomarkers.

The aim of this study was to develop an automatic segmentation method for hip abductor muscles and find their fat fraction associations with early stage hip osteoarthritis (OA) cartilage degeneration biomarkers. This Institutional Review Board approved, Health Insurance Portability and Accountability Act compliant prospective study recruited 61 patients with evidence of hip OA or Femoroacetabular Impingement (FAI). Magnetic resonance (MR) images were acquired for cartilage segmentation, T and T relaxation times computation and grading of cartilage lesion scores. A 3D V-Net (Dice loss, Adam optimizer, learning rate = 1e , batch size = 3) was trained to segment the three muscles (gluteus medius, gluteus minimus, and tensor fascia latae). The V-Net performance was measured using Dice, distance maps between manual and automatic masks, and Bland-Altman plots of the fat fractions and volumes. Associations between muscle fat fraction and T , T relaxation times values were found using voxel based relaxometry (VBR). A p < 0.05 was considered significant. The V-Net had a Dice of 0.90, 0.88, and 0.91 (GMed, GMin, and TFL). The VBR results found associations of fat fraction of all three muscles in early stage OA and FAI patients with T , T relaxation times. Using an automatic, validated segmentation model, the associations derived between OA biomarkers and muscle fat fractions provide insight into early changes that occur in OA, and show that hip abductor muscle fat is associated with markers of cartilage degeneration.