Traditional machine learning for limited angle tomography
- 102 Downloads
The application of traditional machine learning techniques, in the form of regression models based on conventional, “hand-crafted” features, to artifact reduction in limited angle tomography is investigated.
Mean-variation-median (MVM), Laplacian, Hessian, and shift-variant data loss (SVDL) features are extracted from the images reconstructed from limited angle data. The regression models linear regression (LR), multilayer perceptron (MLP), and reduced-error pruning tree (REPTree) are applied to predict artifact images.
REPTree learns artifacts best and reaches the smallest root-mean-square error (RMSE) of 29 HU for the Shepp–Logan phantom in a parallel-beam study. Further experiments demonstrate that the MVM and Hessian features complement each other, whereas the Laplacian feature is redundant in the presence of MVM. In fan-beam, the SVDL features are also beneficial. A preliminary experiment on clinical data in a fan-beam study demonstrates that REPTree can reduce some artifacts for clinical data. However, it is not sufficient as a lot of incorrect pixel intensities still remain in the estimated reconstruction images.
REPTree has the best performance on learning artifacts in limited angle tomography compared with LR and MLP. The features of MVM, Hessian, and SVDL are beneficial for artifact prediction in limited angle tomography. Preliminary experiments on clinical data suggest that the investigation on more features is necessary for clinical applications of REPTree.
KeywordsMachine learning Limited angle tomography Decision tree
Compliance with ethical standards
Conflict of interest
Oliver Taubmann and Guenter Lauritsch are with Siemens Healthcare GmbH, Forchheim, Germany. Yixing Huang is supported by Siemens Healthcare GmbH, Forchheim, Germany.
All data shared in the challenge were fully anonymized. This article does not contain any studies with animals performed by any of the authors.
The clinical data in this paper are from the library of the Low Dose CT Grand Challenge . The library was HIPAAcompliant and built with waiver of informed consent.
The concepts and information presented in this paper are based on research and are not commercially available.
- 7.Huang Y, Taubmann O, Huang X, Lauritsch G, Maier A (2018) Papoulis–Gerchberg algorithms for limited angle tomography using data consistency conditions. Procs CT Meeting, pp 189–192Google Scholar
- 21.Würfl T, Hoffmann M, Christlein V, Breininger K, Huang Y, Unberath M, Maier AK (2018) Deep learning computed tomography: Learning projection-domain weights from image domain in limited angle problems. IEEE Trans Med Imaging 37Google Scholar
- 22.Riess C, Berger M, Wu H, Manhart M, Fahrig R, Maier A (2013) TV or not TV? That is the question. Procs Fully 3D: 341–344Google Scholar
- 23.Hammernik K, Würfl T, Pock T, Maier A (2017) A deep learning architecture for limited-angle computed tomography reconstruction. Procs BVM pp 92–97Google Scholar
- 24.Ronneberger O, Fischer P, Brox T (2015) U-net: convolutional networks for biomedical image segmentation. Procs MICCAI pp 234–241Google Scholar
- 25.Gu J, Ye JC (2017) Multi-scale wavelet domain residual learning for limited-angle CT reconstruction. Procs Fully 3D:443–447Google Scholar
- 26.Zeiler MD, Fergus R (2014) Visualizing and understanding convolutional networks. Eur Conf Comput Vis pp 818–833Google Scholar
- 31.McCollough CH, Bartley AC, Carter RE, Chen B, Drees TA, Edwards P, Holmes DR, Huang AE, Khan F, Leng S, McMillan KL, Michalak GJ, Nunez KM, Yu L, Fletcher JG (2017) Low-dose CT for the detection and classification of metastatic liver lesions: Results of the 2016 low dose CT grand challenge. Med Phys 44(10)Google Scholar
- 32.Frank E, Hall MA, Witten IH (2016) The WEKA workbench. Online appendix for data mining: practical machine learning tools and techniques (Morgan Kaufmann, 2016)Google Scholar