Identification of Degenerative Cartilage Through Use of Multiparametric Discriminant Analysis and the Support Vector Machine

Ping-Chang Lin¹, Richard G. Spencer¹

MRI is increasingly employed as a means of describing cartilage; however, using univariate analysis has limitedly succeeded to classify tissue into normal and degraded groups due to a large degree of overlap in distributed parameter values. We extend the simple arithmetic means procedure by including multivariate model-based discriminant and support vector machine analyses applying on the control puls trypsin or collagenase digested cartilage samples. Results from these three approaches are analyzed according to their sensitivities and specificities. Classification accuracy is markedly improved by either one of these two multivariate procedures, compared to the best uni-parametric MR discriminator.