Abstract #0136

Improving the reliability of pharmacokinetic parameters in dynamic contrast-enhanced MRI in gliomas: Deep learning approach

Kyu Sung Choi¹, Sung-Hye You², Yoseob Han¹, Jong Chul Ye¹, Seung Hong Choi³, and Bumseok Jeong¹

¹Korea Advanced Institute for Science and Technology, Daejeon, Korea, Republic of, ²Korea University College of Medicine, Seoul, Korea, Republic of, ³Seoul National University Hospital, Seoul, Korea, Republic of

AIF_DCE has been known to be sensitive to noise, because of the relatively weak T1 contrast-enhanced MR signal intensity (SI) compared to the T2* SI of DSC-MRI, leading to PK parameters – K_trans, V_e, and _Vp – with low reliability. In this study, we developed a neural network model generating an AIF similar to the AIF obtained from DSC-MRI – AIF_{generated DSC} – and demonstrated that the accuracy and reliability of K_trans and V_e derived from AIF_{generated DSC} can be improved compared to those from AIF_DCE without obtaining DSC-MRI, not leading to an additional deposition of gadolinium in the brain.

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

Join Here