Three-dimensional artificial intelligence-based computed tomography analysis of lower limb muscle volume and fatty degeneration in varus and valgus knee osteoarthritis: a single-center retrospective study.
Authors
Affiliations (9)
Affiliations (9)
- Department of Orthopaedic Surgery, Nara Medical University, 840 Shijo-cho, Kashihara, Nara, 634-8521, Japan. [email protected].
- Department of Orthopaedic Surgery, Nara Medical University, 840 Shijo-cho, Kashihara, Nara, 634-8521, Japan.
- Department of Sports Medicine, Nara Medical University, Kashihara, Japan.
- The Center for Rheumatic Diseases, Nara Medical University, Kashihara, Japan.
- Department of Rehabilitation Medicine, Nara Medical University, Kashihara, Japan.
- Department of Orthopaedic Medical Engineering, Osaka University Graduate School of Medicine, Suita, Japan.
- Department of Orthopaedic Surgery, Ehime University Graduate School of Medicine, Toon, Japan.
- Program of Information and Communication Engineering, Faculty of Engineering, University of Miyazaki, Miyazaki, Japan.
- Division of Information Science, Nara Institute of Science and Technology, Ikoma, Japan.
Abstract
Despite the high prevalence of knee osteoarthritis, the anatomical factors that characterize distinct deformity patterns remain unclear. This study aimed to compare the three-dimensional (3D) lower-limb musculature between varus and valgus knee osteoarthritis (OA) using an artificial intelligence (AI)-based computed tomography (CT) segmentation, evaluating whole-muscle volume, fatty degeneration, and morphological differences to inform targeted prevention, treatment, and rehabilitation strategies. This retrospective study included 75 patients who underwent CT imaging before total knee arthroplasty for end-stage knee OA. Patients were categorized into valgus (femorotibial angle [FTA] ≤ 176°) and varus (FTA > 176°) groups. A validated Bayesian U-Net-based AI segmentation model was used to reconstruct 3D musculature and quantify muscle volume and Hounsfield unit (HU) values. Twenty-three muscle compartments were analyzed. For the vastus medialis and tensor fasciae latae (TFL), intramuscular fat volume was additionally quantified using Mimics to calculate fat percentage. Between-group comparisons were performed using the Mann-Whitney U test, and false discovery rate correction was applied for multiple comparisons. The valgus group (n = 16) demonstrated significantly smaller muscle volumes in the biceps femoris, TFL, and superficial posterior compartment than the varus group (n = 59) after false discovery rate correction. The HU value of the TFL was significantly lower in the valgus group. The vastus medialis showed a lower HU value before correction; however, this difference did not remain significant after false discovery rate correction. Mimics-based fat quantification revealed no significant difference in the fat percentage of the vastus medialis, whereas the TFL showed significantly higher fat infiltration in valgus knees. Correlation analyses using FTA as a continuous variable showed that greater valgus alignment was associated with lower TFL muscle volume, lower TFL HU values, and greater TFL fat infiltration. AI-assisted CT-based 3D analysis revealed distinct muscle morphologies and fatty degeneration patterns in varus and valgus knees of patients with OA. Valgus knees exhibited characteristic reductions in the posterior and lateral muscle groups and increased fatty infiltration of the TFL. These deformity-specific muscular alterations may inform personalized rehabilitation strategies and help elucidate the factors contributing to the progression of OA. Not applicable.