Classification of Cartilage Degradation and Quantification of Matrix Composition Through Multiparametric Support Vector Machine Analysis

Ping-Chang Lin¹, Onyi Irrechukwu¹, Remy Roque¹, Richard G. Spencer¹

Univariate classification, as is implicitly used in analyses of cartilage matrix using MRI parameters, exhibits limited ability to discriminate between control and degraded tissue. In view of these limitations, we undertook a multivariate support vector machine (SVM) analysis of bovine nasal cartilage (BNC) samples with pathomimetic degradation using trypsin and collagenase. Our current results, that the sets (T₁, k_m), (T₁, T₂, k_m) and (T₁, k_m, ADC) exhibit particularly favorable classification properties, are consistent with our previous study, indicating that these parameter combinations may emerge as particularly useful in multivariate cartilage matrix characterization