Generation of abdominal synthetic CTs from 0.35T MR images using generative adversarial networks for MR-only liver radiotherapy

Abstract

Electron density maps must be accurately estimated to achieve valid dose calculation in MR-only radiotherapy. The goal of this study is to assess whether two deep learning models, the conditional generative adversarial network (cGAN) and the cycle-consistent generative adversarial network (cycleGAN), can generate accurate abdominal synthetic CT (sCT) images from 0.35T MR images for MR-only liver radiotherapy.A retrospective study was performed using CT images and 0.35T MR images of 12 patients with liver (n=8) and non-liver abdominal (n=4) cancer. CT images were deformably registered to the corresponding MR images to generate deformed CT (dCT) images for treatment planning. Both cGAN and cycleGAN were trained using MR and dCT transverse slices. Four-fold cross-validation testing was conducted to generate sCT images for all patients. The HU prediction accuracy was evaluated by voxel-wise similarity metric between each dCT and sCT image for all 12 patients. dCT-based and sCT-based dose distributions were compared using gamma and dose-volume histogram (DVH) metric analysis for 8 liver patients. sCTcycleGAN achieved the average mean absolute error (MAE) of 94.1 HU, while sCTcGAN achieved 89.8 HU. In both models, the average gamma passing rates within all volumes of interest were higher than 95% using a 2%, 2 mm criterion, and 99% using a 3%, 3 mm criterion. The average differences in the mean dose and DVH metrics were within ±0.6% for the planning target volume and within ±0.15% for evaluated organs in both models.Results demonstrated that abdominal sCT images generated by both cGAN and cycleGAN achieved accurate dose calculation for 8 liver radiotherapy plans. sCTcGAN images had smaller average MAE and achieved better dose calculation accuracy than sCTcyleGAN images. More abdominal patients will be enrolled in the future to further evaluate two models. Keywords: Generative adversarial network, Synthetic CT, MR-guided radiotherapy planning workflows for pelvic or abdominal cancer radiotherapy (Villeirs et al 2005, Lim et al 2011, Heerkens et al 2017, Mittauer et al 2018. Since there is no direct relationship between MR intensity values and electron densities, the standard MR-guided radiotherapy workflow still requires the acquisition of a CT image for dose calculation. However, registration between CT and MR images for transferring target delineations introduces systematic uncertainties that propagate throughout the treatment (Edmund and Nyholm 2017). Acquiring an additional CT image also increases unwanted radiation exposure, clinical workload, and financial cost (Karlsson et al 2009). MR-only radiotherapy can avoid these downsides.A few methods have been proposed to generate synthetic CT (sCT) images from MR images. These methods include atlas-based methods, voxel-based methods, and hybrid methods (Edmund and Nyholm 2017). In atlas-based methods (Sjölund et al 2015, Dowling et al 2015, the target MR image was first deformably registered to atlas-MR images to acquire deformation vector fields. The acquired vector fi...

Deep learning based synthetic‐CT generation in radiotherapy and PET: A review

“… 76 compared a 2D U‐net trained in a paired manner against a cycle‐GAN trained in an unpaired manner, finding that image similarity was higher with the U‐net. Similarly, two other studies compared 2D‐paired against unpaired GANs, achieving slightly better similarity and lower DD with paired training in the abdomen 75 and H&N 67 . Mixed paired/unpaired training was proposed by Jin et al 89 .…”

“…Four studies compared paired against unpaired. 67,[75][76][77] The 2D networks were the most common over the three categories,being adopted about 61% of the times, 2D+ 6%, 2.5D 10%, and 3D configuration 24%. In some studies, multiple configurations were investigated, for example.…”

“…The 2D networks were the most common over the three categories, being adopted about 61% of the times, 2D+ 6%, 2.5D 10%, and 3D configuration 24%. In some studies, multiple configurations were investigated, for example 75,78,79 . GANs were the most popular architectures (45 times), followed by U‐nets (36) and other CNNs.…”

“…For anatomical regions, as pelvis and brain/H&N, high image similarity and dosimetric accuracy, that is, DDs , can be achieved for photon RT and proton therapy. In region strongly affected by motion, 156,157 for example, abdomen and thorax, the first feasibility studies seem to be promising 72,75,82,112,121 . However, no study proposed the generation of DL‐based 4D sCT yet, as from classical methods 158 .…”

Magnetic resonance-based synthetic computed tomography images generated using generative adversarial networks for nasopharyngeal carcinoma radiotherapy treatment planning

“…Deep learning models have been applied for generating SCT images in radiotherapy planning for the prostate or pelvic [18,27,43], brain [17,20,23,25,30], liver [44][45][46], breast [29], and head and neck cancer [21,24,[47][48][49]. However, for cancers with more complicated treatment sites such as nasopharyngeal carcinoma (NPC), generating SCT images from conventional MR image remains challenging because several air-bearing bone structures near and/or within the nasopharynx.…”