2020/10/19 by Matthew Tivnan, J. Webster Stayman, Tivnan, Matthew +1
Medicine · #Colorectal Cancer Screening and Detection #FOS: Physical sciences #Medical Imaging Techniques and Applications #Medical Physics (physics.med-ph) #Radiomics and Machine Learning in Medical Imaging
paper · pdf · doi:10.48550/arxiv.2010.09685
openalex publication_date 2020/10/19 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Manifold learning using deep neural networks been shown to be an effective\ntool for building sophisticated prior image models that can be applied to noise\nreduction in low-dose CT. We propose a new iterative CT reconstruction\nalgorithm, called Manifold Reconstruction of Differences (MRoD), which combines\nphysical and statistical models with a data-driven prior based on manifold\nlearning. The MRoD algorithm involves estimating a manifold component,\napproximating common features among all patients, and the difference component\nwhich has the freedom to fit the measured data. By applying a\nsparsity-promoting penalty to the difference image rather than a hard\nconstraint to the manifold, the MRoD algorithm is able to reconstruct features\nwhich are not present in the training data. The difference component itself may\nbe independently useful. While the manifold captures typical patient features\n(e.g. healthy anatomy), the difference image highlights patient-specific\nelements (e.g. pathology). In this work, we present a description of an\noptimization framework which combines trained manifold-based modules with\nphysical modules. We present a simulation study using anthropomorphic lung data\nshowing that the MRoD algorithm can both isolate differences between a\nparticular patient and the typical distribution, but also provide significant\nnoise reduction with less bias than a typical penalized likelihood estimator in\ncomposite manifold plus difference reconstructions.\n