Abstract #3501

Can 3D Shape and Textural Analysis Differentiate Liposarcomas from Benign Lipomas?

Rebecca E. Thornhill¹, Greg O. Cron², Mohammad Golfam¹, Gina DiPrimio¹, Adnan Sheikh¹, Eric A. White³, Joel Werier⁴, Mark E. Schweitzer¹

¹Medical Imaging, The Ottawa Hospital, Ottawa, Ontario, Canada; ²Medical Imaging, Ottawa Hospital Research Institute, Ottawa, Ontario, Canada; ³Radiology, University of Southern California, Los Angeles, CA, United States; ⁴Surgery, The Ottawa Hospital, Ottawa, Ontario, Canada

Conventional imaging assessment of lipomatous tumors offers limited accuracy to predict malignant potential. We have identified a number of quantitative morphological and textural features that can be extracted from standard T1-weighted spin echo MR images that are capable of delineating lipomas from liposarcomas with 88% sensitivity, 90% specificity and 89% accuracy. These advanced 3 dimensional shape and textural features may identify a 'signature' pattern associated with highly malignant lipomatous tumors and offer improved diagnosis compared to conventional radiologic methods.