Visual enhancement of Cone‐beam CT by use of CycleGAN

Abstract

Purpose Cone‐beam computed tomography (CBCT) offers advantages over conventional fan‐beam CT in that it requires a shorter time and less exposure to obtain images. However, CBCT images suffer from low soft‐tissue contrast, noise, and artifacts compared to conventional fan‐beam CT images. Therefore, it is essential to improve the image quality of CBCT. Methods In this paper, we propose a synthetic approach to translate CBCT images with deep neural networks. Our method requires only unpaired and unaligned CBCT images and planning fan‐beam CT (PlanCT) images for training. The CBCT images and PlanCT images may be obtained from other patients as long as they are acquired with the same scanner settings. Once trained, three‐dimensionally reconstructed CBCT images can be directly translated into high‐quality PlanCT‐like images. Results We demonstrate the effectiveness of our method with images obtained from 20 prostate patients, and provide a statistical and visual comparison. The image quality of the translated images shows substantial improvement in voxel values, spatial uniformity, and artifact suppression compared to those of the original CBCT. The anatomical structures of the original CBCT images were also well preserved in the translated images. Conclusions Our method produces visually PlanCT‐like images from CBCT images while preserving anatomical structures.

Evaluation of a cycle-generative adversarial network-based cone-beam CT to synthetic CT conversion algorithm for adaptive radiation therapy

“…Related cycle-GAN-based sCT studies reported similar values for the ME/MAE of the total body contour with − 3/ 16.1HU [21] and − 6/87HU [33]. The analysis of Kida et al [30] which compared cycle-GAN generated sCT to deformed CT yielded mean HU values for the prostate and bladder of 19HU and 4HU. However, a direct comparison remains difficult since all referenced studies differed from the presented work with regard to the size of the patient collective, the manufacturer-specific image acquisition parameters, the detailed architecture of the GAN and the training of the body site-specific models.…”

“…Regarding target volume dose metrics, two related studies reported dosimetric differences between the sCT and CT with a comparable accuracy to our study. A V 95 of 0.2-0.7% for the H&N body site [35] and an absolute difference of D 98 , D 50 and D 2 of − 1 to 0 Gy for the pelvic body site [30] were obtained, respectively. For the thoracic body region, D 98 showed the largest deviations and for the D 50 and D 2 outliers were observed which can be attributed to remaining uncertainties due to the breathing pattern.…”

“…In order to address these issues, additional modifications of the original cycle-GAN network were implemented by Elekta and are part of ADMIRE. Firstly, different to the unpaired training mechanism adopted in several previous research works [22,29,30], the training strategy in this sCT module is based on paired CBCT and pCT. The paired pCT were obtained by applying a DIR in ADMIRE on the original pCT.…”

Comparison of CBCT‐based dose calculation methods in head and neck cancer radiotherapy: from Hounsfield unit to density calibration curve to deep learning

“…Studies using DL for pCT generation from CBCT are scarce for brain, 55 H&N, [13][14][15] pancreas, 38 or prostate cancer. [35][36][37]55,56 The studies involved an imaging analysis (pCT vs. reference CT), but only half of them evaluated the dose accuracy. Among the three H&N studies using DL for pCT generation from CBCT, [13][14][15] one involved training a U-Net neural network on 50 coregistered CBCT/CT images and performing a test based on data from 10 patients.…”

“…In fact, pCT generated from CBCT are used to monitor delivered doses or to estimate the cumulative delivered dose during the treatment course in the context of dose‐guided adaptive radiotherapy. Studies using DL for pCT generation from CBCT are scarce for brain, 55 H&N, 13–15 pancreas, 38 or prostate cancer 35–37,55,56 . The studies involved an imaging analysis (pCT vs. reference CT), but only half of them evaluated the dose accuracy.…”