AUTHOR=Tahri Safaa , Texier Blanche , Nunes Jean-Claude , Hemon Cédric , Lekieffre Pauline , Collot Emma , Chourak Hilda , Le Guevelou Jennifer , Greer Peter , Dowling Jason , Acosta Oscar , Bessieres Igor , Marage Louis , Boue-Rafle Adrien , De Crevoisier Renaud , Lafond Caroline , Barateau Anaïs 

TITLE=A deep learning model to generate synthetic CT for prostate MR-only radiotherapy dose planning: a multicenter study

JOURNAL=Frontiers in Oncology

VOLUME=Volume 13 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2023.1279750

DOI=10.3389/fonc.2023.1279750

ISSN=2234-943X

ABSTRACT=Introduction: For radiotherapy based solely on MRI, generating synthetic CT (sCT) from MRI is essential for dose calculation. The use of deep learning (DL) methods to generate sCT from MRI have shown encouraging results if the MRI images used for training of the deep learning network and the MRI images for sCT generation come from the same MRI device. The objective of this study was to create and evaluate a generic DL model capable of generating sCTs from various MRI devices for prostate radiotherapy.
Material and methods: Ninety patients from three centers (30 CT–MR prostate pairs/center) underwent treatment using VMAT for prostate cancer (Pca) (60 Gy in 20 fractions). T2 MRI images were acquired in addition to CT images for treatment planning. The DL model was a 2D supervised conditional GAN (Pix2Pix). Patient images underwent preprocessing steps including non-rigidly registration. Seven different supervised models were trained, incorporating patients from 1, 2, or 3 centers. Each model was trained on 24 CT–MR prostate pairs. A generic model was trained using patients from all three centers. To compare sCT and CT, the mean absolute error in Hounsfield units was calculated. For dose analysis, mean dose differences of D99% for CTV, V95% for PTV, Dmax for rectum and bladder, and gamma analysis were calculated. Furthermore, Wilcoxon tests were performed to compare the image and dose results obtained with the generic model to those with the other trained models.
Results: Considering the image results for the entire pelvis, when the data used for the test comes from the same center as the data used for training, the results were non-significantly different from the generic model. Absolute dose differences were less than 1 Gy for the CTV D99% for every trained model and center. The gamma analysis results showed non-significant differences between the generic and monocentric models.
Conclusion: The accuracy of sCT, in terms of image and dose, is equivalent whether MRI images are generated using the generic model or the monocentric model. The generic model, using only 8 MRI-CT pairs per center, offers robust sCT generation, facilitating PCa MRI-only radiotherapy for routine clinical use.