Meeting Banner
Abstract #1107

Mapping Cartilage Degradation through Support Vector Machine Probabilistic Classification

Ping-Chang Lin1, Onyi Irrechukwu1, Richard G. Spencer1

1National Institute on Aging, National Institutes of Health, Baltimore, MD, United States

A major limitation of MRI approaches for detection of early osteoarthritis is that individual MRI parameters exhibit substantial overlap between different stages of degradation. To overcome this, we are developing support vector machine (SVM)-based fuzzy classification. We present results obtained on bovine nasal cartilage subjected to pathomimetic enzymatic degradation. SVM analysis was performed on combinations of the parameter set {T1, T2, km, ADC}. Probabilistic maps resulting from the classification procedure represent maps of degradation, with values ranging from zero to unity for assignment to non-degraded status. These maps provide a substantial improvement over univariate MR maps for defining cartilage degradation.