Abstract #3127

Combined MRI Texture & Shape Analysis for the Prediction of Biologic Aggressiveness in Musculoskeletal Neoplasms

Rebecca E. Thornhill¹, Greg O. Cron¹, Ian Cameron¹, Adnan Sheikh¹, Gina Di Primio¹, Joel Werier¹, Mark E. Schweitzer¹, Jing Zhang², Xiao Guang Cheng²

¹The Ottawa Hospital, Ottawa, Ontario, Canada; ²Beijing Ji Shui Tan Hospital, Beijing, China, People's Republic of

Musculoskeletal tumors comprise a diverse group of uncommon neoplasms. Conventional contrast-enhanced MRI techniques are inadequate for distinguishing malignant from benign masses. Quantitative dynamic contrast-enhanced (DCE-)MRI techniques followed by tracer kinetic modeling may offer improved diagnostic accuracy, but requires descriptors that can capture the spatial extent (shape) and complexity (texture) of tracer kinetic parameters like K^trans. We retrospectively evaluated both textural and shape features extracted from K^trans maps obtained from 86 patients with biopsy confirmed musculoskeletal tumors. This combined analysis achieved a sensitivity, specificity, and accuracy of 75%, 79%, and 77%, respectively for the discrimination of benign from malignant masses.