Click here to

Session: Imaging for Particle Therapy [Return to Session]

Synthetic Dual-Energy CT for MRI-Based Proton Therapy Treatment Planning Using Label-GAN

R Liu1*, Y Lei2, T Wang3, J Zhou4, J Roper5, L Lin6, M McDonald7, J Bradley8, T Liu9, X Yang10, (1) Emory University, Atlanta, GA, (2) Emory Univ, Atlanta, GA, (3) Emory University, Atlanta, GA, (4) Emory University, Atlanta, GA, (5) Winship Cancer Institute of Emory University, Atlanta, GA, (6) Emory Proton Therapy Center, Atlanta, GA, (7) Emory University, Atlanta, GA, (8) Emory University School Of Medicine, ,,(9) Emory Univ, Atlanta, GA, (10) Emory University, Atlanta, GA


TU-B-TRACK 6-6 (Tuesday, 7/27/2021) 11:30 AM - 12:30 PM [Eastern Time (GMT-4)]

Purpose: MRI-only treatment planning is highly desirable in current proton radiation therapy workflow due to its appealing advantages such as bypassing MR-CT co-registration, avoiding x-ray CT exposure dose and reduced medical cost. However, MRI alone cannot provide stopping power ratio (SPR) information for dose calculations. Given that dual energy CT (DECT) can estimate SPR with higher accuracy than conventional single energy CT (SECT), we propose a deep learning-based method in this study to generate synthetic DECT (sDECT) from MRI to calculate SPR.

Methods: Since the contrast difference between high-energy and low-energy CT is important, and in order to accurately model this difference, we propose a novel label generative adversarial network (label-GAN)-based model which can not only discriminate the realism of sDECT but also differentiate high-energy CT (HECT) and low-energy CT (LECT) from DECT. A cohort of 57 head-and-neck cancer patients with DECT and MRI pairs were used to validate the performance of the proposed framework. The results of sDECT and its derived SPR maps were compared with clinical DECT and the corresponding SPR, respectively.

Results: The mean absolute error (MAE) for synthetic LECT and HECT were 79.98±18.11 HU and 80.15±16.27 HU, respectively. The corresponding SPR maps generated from sDECT showed a normalized mean absolute error (NMAE) as 5.22%±1.23%. By comparing with the traditional Cycle GANs, our proposed method significantly improves the accuracy of sDECT.

Conclusion: We developed a deep learning-based method to generate sDECT from MRI for proton radiotherapy. The results indicate that on our dataset, the synthetic DECT image form MRI is close to planning DECT, and thus shows promising potential for generating SPR maps for proton therapy. The image similarity and SPR agreement between SDECT and DECT warrants further study and development of an MRI-only workflow for MRI-based proton radiotherapy.



    Not Applicable / None Entered.


    Not Applicable / None Entered.

    Contact Email