Abstract #0685

Automated Imaging Classification Based On Volumetric Analysis: Application On Primary Progressive Aphasia

Andreia Vasconcellos Faria^1,2, Kyrana Tsapkini³, Jennifer Crinion⁴, Hangyi Jiang¹, Xin Li¹, Kenichi Oishi¹, Peter van Zijl¹, Michael Miller⁵, Argye Hillis³, Susumu Mori¹

¹Radiology, Johns Hopkins University, Baltimore, MD, USA; ²Radiology, State University of Campinas, Campinas, SP, Brazil; ³Neurology, Johns Hopkins University, Baltimore, MD, USA; ⁴Institute of Cognitive Neuroscience, University College London; ⁵Biomedical Engineering, Johns Hopkins University, Baltimore, MD, USA

Based on large-deformation diffeomorphic metric mapping and Atlas-based analysis, we developed a method to capture the anatomical features and classifying primary progressive aphasia (PPA) patients. Principal component analysis, multivariate techniques and predictive modeling were applied to a 32 PPA patients and 27 controls. The variables used to create and test the classification model were selected among the volumes of the 211 regions obtained from the automated 3D segmentation. The percentage of correct classification was 83% after two-level cross-validation. The results from this automated quantitative analysis can be user-friendly displayed and this method can be applied to routine clinical practice