Quantitative Sonographic Assessment of Synovitis and its Relationship to Joint Damage and Pain in Knee Osteoarthritis
Abstract
Osteoarthritis (OA) is a chronic condition affecting synovial joints, manifesting as pain, stiffness, and disability. Substantial evidence links inflammation of the synovial membrane (synovitis) to joint damage and knee pain, motivating the need for non-invasive and reproducible methods of synovitis assessment. The purpose of this thesis is to develop and test a new method of analyzing sonographic images of knee synovitis and, by integrating machine learning approaches, investigate the relationships between synovitis, joint damage, and OA-related knee pain.
Chapter two reviews imaging-based methods of synovitis assessment in knee OA, highlighting magnetic resonance and ultrasound as the most commonly used modalities. Various semi-quantitative and quantitative methods exist but lack consensus on a gold standard. Inconsistent correlations with knee pain and microscopic synovitis reinforce that synovitis represents both active inflammation and synovial remodeling.
Chapter three explores the measurement properties of a novel image analysis methodology for sonographic images of knee synovitis. This method separately quantifies effusion and synovitis without intravenous contrast, correlates strongly with existing measures, and is amenable to automation/pattern analysis using machine-learning (ML) methods.
Chapter four investigates the relationships between quantitative measures of synovitis, joint damage, and pain among patients with active early and late-stage knee OA. Synovial hyperplasia is more strongly associated with knee pain than effusion in early and late-stage disease. Inflammation in knee OA increases with joint damage in early-stage, but not in late-stage disease, which supports a theory of joint failure secondary to maladaptive synovial remodeling. These results have significant implications on our understanding of synovitis as an outcome measure and marker of joint homeostasis.
Chapter five employs ML to analyze the clustering behaviors of features representing radiographic OA severity, knee-specific pain, and synovitis. Three clinical phenotypes of knee OA are described- “early disease”, “active disease”, and “total joint failure”. Patients in the total joint failure group exhibit low synovitis alongside high pain and joint damage, with increased rates of systemic metabolic dysregulation.
Overall, these findings provide new insights into the complex relationships between inflammation, joint damage, and pain in knee OA, and open new avenues for ML in the processing and analysis of ultrasound-synovitis images.