Abstract #2326

A deep neural network based model for treatment response prediction using longitudinal diffusion MRI

Yu Gao^1,2, Vahid Ghodrati^1,2, Anusha Kalbasi³, Jie Fu^2,3, Dan Ruan^2,3, Minsong Cao^2,3, Chenyang Wang³, Fritz C. Eilber⁴, Nicholas Bernthal⁵, Susan Bukata⁵, Sarah M. Dry⁶, Scott D. Nelson⁶, Mitchell Kamrava⁷, John Lewis^2,3, Daniel A. Low^2,3, Michael Steinberg³, Peng Hu^1,2, and Yingli Yang^2,3

¹Department of Radiological Sciences, University of California, Los Angeles, Los Angeles, CA, United States, ²Physics and Biology in Medicine IDP, University of California, Los Angeles, Los Angeles, CA, United States, ³Department of Radiation Oncology, University of California, Los Angeles, Los Angeles, CA, United States, ⁴Division of Surgical Oncology, Department of Surgery, University of California, Los Angeles, Los Angeles, CA, United States, ⁵Department of Orthopaedic Surgery, University of California, Los Angeles, Los Angeles, CA, United States, ⁶Department of Pathology, University of California, Los Angeles, Los Angeles, CA, United States, ⁷Radiation Oncology, Cedars-Sinai Medical Center, Los Angeles, CA, United States

A deep neural network based model was proposed to predict post-radiotherapy treatment effect score for localized soft tissue sarcoma patient using longitudinal diffusion MRI. Diffusion images were acquired three times throughout patient’s hypofractionated radiotherapy treatment. A convolutional neural network was constructed to learn the most useful spatial features from the tumor ADC maps at each time point, which is then fed into a recurrent neural network to exploit the temporal information between the extracted features. Excellent prediction performance of 97.4% accuracy on slice-based classification, and 95% accuracy on patient-based classification were achieved on independent test sets.

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

Join Here