Computer Graphics Forum · 2018 · 113 citations · 20 references
Computed TomographyImage ReconstructionEngineeringX-ray ImagingImage AnalysisImage-based ModelingStereoscopic RenderingComputational ImagingX‐ray ImageRadiologyHealth SciencesMedical ImagingSingle‐image TomographyMedical Image ComputingDeep LearningVolume RenderingComputer VisionDense ReconstructionBiomedical ImagingConvolutional Neural NetworksTomography3D Imaging
Reconstructing a 3D volume from a single 2D x‑ray is difficult because many different 3D shapes can produce the same projection. The authors aim to develop a method that can recover 3D cranial volumes from a single 2D x‑ray to enable applications such as stereoscopic rendering and re‑rendering with altered illumination or pose. They train a two‑stage convolutional neural network that first predicts a coarse fixed‑resolution volume and then refines it by fusing the input x‑ray to produce a high‑resolution reconstruction, using a newly created dataset of nearly half a million simulated x‑rays from 175 mammalian species. The approach demonstrates that deep learning‑based CNNs can successfully solve single‑image tomography, outperforming prior methods and validated through comparisons, a user study, and real‑x‑ray tests.
Abstract As many different 3D volumes could produce the same 2D x‐ray image, inverting this process is challenging. We show that recent deep learning‐based convolutional neural networks can solve this task. As the main challenge in learning is the sheer amount of data created when extending the 2D image into a 3D volume, we suggest firstly to learn a coarse, fixed‐resolution volume which is then fused in a second step with the input x‐ray into a high‐resolution volume. To train and validate our approach we introduce a new dataset that comprises of close to half a million computer‐simulated 2D x‐ray images of 3D volumes scanned from 175 mammalian species. Future applications of our approach include stereoscopic rendering of legacy x‐ray images, re‐rendering of x‐rays including changes of illumination, view pose or geometry. Our evaluation includes comparison to previous tomography work, previous learning methods using our data, a user study and application to a set of real x‐rays.
20
Deep Residual Learning for Image Recognition
Kaiming He, Xiangyu Zhang, Shaoqing Ren et al. · 2016 · 214.9K citations · Full text
Image Classification, Deep Neural Networks, Machine Vision +14
Image quality assessment: from error visibility to structural similarity
Zhou Wang, Alan C. Bovik, Hamid R. Sheikh et al. · IEEE Transactions on Image Processing · 2004 · 54.1K citations
Learning representations by back-propagating errors
David E. Rumelhart, Geoffrey E. Hinton, Ronald J. Williams · Nature · 1986 · 29.7K citations
Yangqing Jia, Evan Shelhamer, Jeff Donahue et al. · 2014 · 11.1K citations
Convolutional Neural Network, Machine Vision, Machine Learning +14