Abstract #5168

Differentiation of Pulmonary From Non-Pulmonary Spine Metastases Using Conventional DCE Kinetic Analysis and Machine Learning

Ning Lang¹, Yang Zhang², Enlong Zhang¹, Jiahui Zhang¹, Daniel Chow², Peter Chang³, Melissa Khy², Hon J. Yu², Huishu Yuan¹, and Min-Ying Su²

¹Department of Radiology, Peking University Third Hospital, Beijing, China, ²Department of Radiological Sciences, University of California, Irvine, CA, United States, ³Department of Radiology, University of California, San Francisco, CA, United States

DCE-MRI of the spine was analyzed to differentiate metastasis from lung cancer (N=30) and other tumors (N=31, 9 breast, 6 prostate, 7 thyroid, 6 liver, 3 kidney). Using DCE parameters measured from the tumor ROI, CHAID decision tree classification selected the wash-out slope of -6.6% and wash-in SE of 98% as thresholds, which could achieve diagnostic accuracy of 0.79. In machine learning, the enhanced tumor on DCE image was segmented automatically by using the normalized cut algorithm. The Convolutional Long Short Term Memory (CLSTM) network with all 12 sets of DCE images as the input could yield accuracy of 0.75-0.84.

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

Join Here