Meeting Banner
Abstract #3297

DTI of Articular Cartilage Can Predict Early Cartilage Damage as Assessed by Histopathology

MAGNA25Jose G. Raya1, Gerd Melkus2, Silvia Adam-Neumair3, Elisabeth Muetzel3, Maximilian F. Reiser3, Peter M. Jakob4, Thorsten Kirsch, Christian Glaser

1Radiology, New York University Langone Medical Center, New York, United States; 2University of California San francisco; 3University of Munich; 4University of Wuerzburg

The aim of this work was to investigate the value of DTI of articular cartilage as predictor for early cartilage damage as assessed by histopathology. 41 cartilage-on-bone samples were examined at 17.6T. After MRI samples underwent histology with safranin-O for histopathology assessment of OA (OARSI score). The prognostic value of DTI was analyzed with logistic regression. Performance of DTI was assessed with ROC-curve analysis and the prediction error with 10-flod cross-validation. DTI demonstrated very good prediction performance (Sensitivity=92.9%, Specificity=77.8%) and a correct classification of more than 3 of 4 samples (78%) in a collective with low OARSI score (12).

Keywords

able according acquisition advance anisotropy articular assess assessed assessment bandwidth best birdcage bivariate blue cartilage central characteristic chosen circles classification classified classify coefficients coil collective combination confidence constituents correctly cross curve cylindrical damage death degeneration degradation deviance deviation diagnosis diagnostic diffusion diffusivity drilled earliest early equally erosion error evaluated even every evidence examined examples fitted fold fraction fractional good goodness grade grades greater harvested healthy highest histology hours human improved in vivo included incorrectly indicate integrity interval invest investigate investigated involved joint kirsch logistic maps matrix medical model models nonparametric odds operating osteoarthritis outliers overall pair patellae performance peter plays plot predict prediction predictor predictors predominantly process prognostic proposed protocol radiology receiver recently regression relationship remainder repeated represent residuals resolution role samples scanner score scores sectional segmented sensitive sensitivity several sided significance significantly slice specificity spin split squares staining stars statistical summarized superficial table tensor trend underwent univariate unpaired validation whole wise zero