My project within the KNEEMO ITN focuses on the Role of Thigh Muscle Cross-Sectional Areas and Fat Tissue for Structural Progression of Knee Osteoarthritis.
Thigh muscle weakness has been suggested to be associated with pain and functional limitations in knee osteoarthritis (OA), but strength measurements are difficult to obtain in the presence of knee pain and could be biased by the pain experienced during the strength measurement itself. The MRI-based analysis of quadriceps cross-sectional areas might therefore be more objective and has been reported to be more sensitive to change than maximal isometric extensor strength. However, manual segmentation of the muscle groups and other structures in thigh MRI is time-consuming. Therefore, we want to achieve an automated segmentation of each anatomical tissue group (subcutaneous fat, femur, medulla, and the muscle groups as quadriceps, flexors, adductors and sartorius) in axial thigh MRIs.
Developing a Software Solution for Automated Thigh Muscle and Adipose Tissue Segmentation using an Active Shape Model
Since quantitative analysis of MR images may be affected by image artefacts and image distortion, a memory based algorithms, the active shape model, will be explored. The active shape model represents each muscle group using a set of points. The training with a dataset allows us to calculate a mean shape for the anatomical groups and generates realistic new shapes similar to ones in training dataset.
The active shape model is trained by previously segmented axial non–fat-suppressed T1 weighted Magnetic Resonance Images of thigh cross- sectional areas. The shape model is computing the mean shape, using manually marked anatomical points as major landmarks and automatic calculated points in each anatomical object.
Project 2: Methods
The Role of Thigh Muscle Cross-Sectional Areas and Fat Tissue for Structural Progression of Knee Osteoarthritis
As a second task in this PhD project, the automated segmentation software will be applied to a well-defined subsample of participants from the Osteoarthritis Initiative (OAI), a large publicly accessible multi-centre longitudinal cohort study of knee OA in the United States. The participants have been specifically selected by the OAI coordinating center as “core imaging cohort”. The participants in this subsample were required to show a KLG 2/3 and frequent pain in one of their knees (target knee) at baseline and received additional image readings by several groups that were funded by the OAI. The results of this series of studies will enable automated and efficient assessment of thigh muscle and adipose tissue morphology to be completed reliably and validly to advance the understanding of the pathogenesis of knee osteoarthritis.
Impact and Dissemination
The development of automated segmentation software will potentially save many resources and associated costs currently expended for imaging analysis. Efficiently identifying muscle deficits from large research cohorts may provide avenues for targeted therapy. The potential research and clinical use of the automated software will be established through analyses of knee OA progression from the OAI database.