MRI-only brain radiotherapy: Assessing the dosimetric accuracy of synthetic CT images generated using a deep learning approach

• Published in: Radiotherapy and oncology Vol. 136; pp. 56 - 63
• Main Authors: Kazemifar, Samaneh, McGuire, Sarah, Timmerman, Robert, Wardak, Zabi, Nguyen, Dan, Park, Yang, Jiang, Steve, Owrangi, Amir
• Format: Journal Article
• Language: English
• Published: Ireland Elsevier B.V 01.07.2019
• Subjects: Bone and Bones - diagnostic imaging; Brain MRI; Brain Neoplasms - diagnostic imaging; Brain Neoplasms - radiotherapy; Deep Learning; GAN structure; Humans; Magnetic Resonance Imaging - methods; Mutual information; Radiotherapy Dosage; Radiotherapy Planning, Computer-Assisted - methods; Radiotherapy, Image-Guided - methods; Radiotherapy, Intensity-Modulated - methods; Synthetic CT; Tomography, X-Ray Computed - methods; Deep learning; Brain MRI; Synthetic CT; GAN structure; Mutual information
• ISSN: 0167-8140; 1879-0887; 1879-0887
• DOI: 10.1016/j.radonc.2019.03.026

•CT and MR images from 77 brain patients who underwent brain tumor radiotherapy are used in this work.•This work demonstrated an accurate and reproducible synthetic CT generation using a GAN model that can distinguish between air and bone regions.•The developed model can reduce radiation dose, patient time, and imaging costs image related to CT simulation and may be a step toward the implementation brain MRI-only radiotherapy. This study assessed the dosimetric accuracy of synthetic CT images generated from magnetic resonance imaging (MRI) data for focal brain radiation therapy, using a deep learning approach. We conducted a study in 77 patients with brain tumors who had undergone both MRI and computed tomography (CT) imaging as part of their simulation for external beam treatment planning. We designed a generative adversarial network (GAN) to generate synthetic CT images from MRI images. We used Mutual Information (MI) as the loss function in the generator to overcome the misalignment between MRI and CT images (unregistered data). The model was trained using all MRI slices with corresponding CT slices from each training subject’s MRI/CT pair. The proposed GAN method produced an average mean absolute error (MAE) of 47.2 ± 11.0 HU over 5-fold cross validation. The overall mean Dice similarity coefficient between CT and synthetic CT images was 80% ± 6% in bone for all test data. Though training a GAN model may take several hours, the model only needs to be trained once. Generating a complete synthetic CT volume for each new patient MRI volume using a trained GAN model took only one second. The GAN model we developed produced highly accurate synthetic CT images from conventional, single-sequence MRI images in seconds. Our proposed method has strong potential to perform well in a clinical workflow for MRI-only brain treatment planning.