Abstract #3224

Sparsity based machine learning algorithms for oxygen extraction fraction mapping

Junghun Cho¹, Hae-Yeoun Lee², jinwei Zhang¹, Pascal Spincemaille³, Hang Zhang⁴, Simon Hubertus⁵, Yan Wen¹, Ramin Jafari¹, Shun Zhang³, Thanh Nguyen³, Ajay Gupta³, and Yi Wang^1,3

¹Biomedical Engineering, Cornell University, New York, NY, United States, ²Computer Software Engineering, Kumoh National Institute of Technology, Gumi, Republic of Korea, ³Radiology, Weill Cornell Medical College, New York, NY, United States, ⁴Electrical and Computer Engineering, Cornell University, New York, NY, United States, ⁵Computer Assisted Clinical Medince, Heidelberg University, Mannheim, Germany

In this work, dictionary and deep learning based algorithms are developed that take advantage of sparse signal representations to improve the accuracy and speed of oxygen extraction fraction (OEF) mapping based on the QSM+qBOLD (QQ) modeling of multi-echo gradient echo data without vascular challenge. The developed dictionary learning (QQ-DL) and deep neural network (QQ-NET) algorithms are significantly faster and provide more accurate OEF maps in simulation than a current algorithm based on cluster analysis of time evolution (CAT). In ischemic stroke patients, QQ-DL and QQ-NET show OEF maps that are consistent with DWI-defined lesions.

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

Join Here