Abstract #3181

MR analysis of thigh muscle myopathy using texture features and supervised machine learning

Hon J Yu¹, Saya Horiuchi^1,2, Toshimi Tando¹, Vincent J Caiozzo³, Virginia E Kimonis⁴, and Hiroshi Yoshioka¹

¹Radiological Sciences, University of California, Irvine, Orange, CA, United States, ²Radiology, St. Luke's International Hospital, Tokyo, Japan, ³Department of Orthopaedics, Physiology & Biophysics, University of California, Irvine, Irvine, CA, United States, ⁴Division of Genetic and Genomic Medicine, Department of Pediatrics, University of California, Irvine, Irvine, CA, United States

This study evaluates texture features to demonstrate their relationship with muscle classification based on a 5-grade scale and their value as classifier when trained in supervised machine-learning framework. The results suggest the texture features capture various image characteristics that are likely utilized during manual muscle classification by human being and can correctly predict with up to 81% accuracy when properly trained in supervised machine-learning setting. A further study with bigger data size would be necessary to fully examine such classification model and also to look into the possibility of selecting subset of features to make such an approach more practical.

This abstract and the presentation materials are available to members only; a login is required.

Join Here