Keywords:
spectral CT, material decomposition, CNN, deep learning
Material decomposition is an important application of spectral Computed Tomography (CT). However, traditional post-processing material decomposition algorithms are based on voxel-local of reconstruction image, which leads two problems. First, ignoring the beam hardening problem. Second, not taking advantage of priori knowledge well. In order to solve these two problems processing spatial specificity, inspired by the work, we import Deep Learning technique to deal with multi-material decomposition. After been trained by plenty of samples, Deep Learning network can break the limit of voxel, reducing the influence of beam hardening and modeling the human body. We build a Convolutional Neural Network (CNN) which is a simplified version of VGG16 net, and simulate some reconstruction data of spectral CT to train the network. Compared to the results of solving linear equations, the CNN method turns out to work much better in the test samples. As the conclusion, we think CNN is useful in the multi-material decomposition, but there still remains many researches to work, such as the source of training data, the balance between the priori knowledge and measurement. On the other hand, this work focuses on post-processing, but using Deep Learning to deal with the pre-processing multi-material decomposition is also a potential direction.
- Zhengyang Chen
- Department of Engineering Physics, Tsinghua University, Beijing, 100084, China & Key Laboratory of Particle & Radiation Imaging (Tsinghua University), Ministry of Education, 100084, China
- Liang Li
- Department of Engineering Physics, Tsinghua University, Beijing, 100084, China & Key Laboratory of Particle & Radiation Imaging (Tsinghua University), Ministry of Education, 100084, China
- Long Y, Fessler J A. Multi-material decomposition using statistical image reconstruction for spectral CT[J]. IEEE transactions on medical imaging, 2014, 33(8): 1614-1626.
- Patino M, Prochowski A, Agrawal M D, et al. Material separation using dual-energy CT: current and emerging applications[J]. Radiographics, 2016, 36(4): 1087-1105.
- Mendonça P R S, Lamb P, Sahani D V. A flexible method for multi-material decomposition of dual-energy CT images[J]. IEEE transactions on medical imaging, 2014, 33(1): 99-116.
- Wang G. A perspective on deep imaging[J]. IEEE Access, 2016, 4: 8914-8924.
- Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:1409.1556, 2014.
- Schmidhuber J. Deep learning in neural networks: An overview[J]. Neural Networks, 2015, 61: 85-117.
- He K, Zhang X, Ren S. Sun J. Deep residual learning for image recognition[C]. CVPR, 2016.
- Li L, Chen Z, Cong W, et al. Spectral CT modeling and reconstruction with hybrid detectors in dynamic-threshold-based counting and integrating modes[J]. IEEE transactions on medical imaging, 2015, 34(3):716-728.
- Sidky E Y, Kao C M, Pan X. Accurate image reconstruction from few-views and limited-angle data in divergent-beam CT[J]. Journal of X-ray Science and Technology, 2006, 14(2): 119-139.