Image quality improvement in cone-beam CT using deep learning

We propose a learning method to generate corrected CBCT (CCBCT) images with the goal of improving the image quality and clinical utility of on-board CBCT. The proposed method integrated a residual block concept into a cyclegenerative adversarial network (cycle-GAN) framework, which is named as Res-cycle GAN in this study. Compared with a GAN, a cycle-GAN includes an inverse transformation from CBCT to CT images, which could further constrain the learning model. A fully convolution neural network (FCN) with residual block is used in generator to enable end-toend transformation. A FCN is used in discriminator to discriminate from planning CT (ground truth) and correction CBCT (CCBCT) generated by the generator. This proposed algorithm was evaluated using 12 sets of patient data with CBCT and CT images. The mean absolute error (MAE), peak signal-to-noise ratio (PSNR), normalized cross correlation (NCC) indexes and spatial non-uniformity (SNU) in the selected regions of interests (ROIs) were used to quantify the correction accuracy of the proposed algorithm. Overall, the MAE, PSNR, NCC and SNU were 20.8±3.4 HU, 32. 8±1.5 dB, 0.986±0.004 and 1.7±3.6%. We have developed a novel deep learning-based method to generate CCBCT with a high image quality. The proposed method increases on-board CBCT image quality, making it comparable to that of the planning CT. With further evaluation and clinical implementation, this method could lead to quantitative adaptive radiotherapy.

[1]  Yang Lei,et al.  Learning‐based CBCT correction using alternating random forest based on auto‐context model , 2018, Medical physics.

[2]  A Dhabaan,et al.  WE-AB-207A-07: A Planning CT-Guided Scatter Artifact Correction Method for CBCT Images. , 2016, Medical physics.

[3]  Yang Lei,et al.  Machine-learning based classification of glioblastoma using dynamic susceptibility enhanced MR image , 2019, Medical Imaging.

[4]  Zhengyang Zhou,et al.  Automated segmentation of the parotid gland based on atlas registration and machine learning: a longitudinal MRI study in head-and-neck radiation therapy. , 2014, International journal of radiation oncology, biology, physics.

[5]  Xiangyang Tang,et al.  A patch-based CBCT scatter artifact correction using prior CT , 2017, Medical Imaging.

[6]  Lei Zhu,et al.  Quantitative cone-beam CT imaging in radiation therapy using planning CT as a prior: first patient studies. , 2012, Medical physics.

[7]  Jian Sun,et al.  Identity Mappings in Deep Residual Networks , 2016, ECCV.

[8]  Lei Zhu,et al.  Image-domain non-uniformity correction for cone-beam CT , 2017, 2017 IEEE 14th International Symposium on Biomedical Imaging (ISBI 2017).

[9]  Ming Dong,et al.  Generating synthetic CTs from magnetic resonance images using generative adversarial networks , 2018, Medical physics.

[10]  Tian Liu,et al.  A denoising algorithm for CT image using low-rank sparse coding , 2018, Medical Imaging.

[11]  Yang Lei,et al.  ePoster SessionsA Learning-Based Approach to Derive Electron Density from Anatomical MRI for Radiation Therapy Treatment Planning , 2017 .

[12]  T. Zhu.  Overview of X-ray Scatter in Cone-beam Computed Tomography and Its Correction Methods , 2010 .

[13]  Alexei A. Efros,et al.  Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks , 2017, 2017 IEEE International Conference on Computer Vision (ICCV).

[14]  Hideaki Endoh,et al.  Spatial non-uniformity of excitation-contraction coupling can enhance arrhythmogenic-delayed afterdepolarizations in rat cardiac muscle. , 2008, Cardiovascular research.