Figure 1 | Dynamical machine learning approach for limited-angle tomography (IMAGE)

Light Publishing Center, Changchun Institute of Optics, Fine Mechanics And Physics, CAS

Figure 1 | Dynamical machine learning approach for limited-angle tomography

Caption

(a) A schematic on the limited-angle tomography. Each illumination angle in an angular axis corresponds to a time step in an analogous temporal axis. (b) An optical apparatus used for experiments under the strong scattering condition. (c) Qualitative comparison on reconstructions of conventional inverse algorithms (FBP: filtered backprojection, FBP + TV: TV-regularized filtered backprojection; TwIST) and the proposed algorithm based on recurrent neural network (RNN). Each column shows a two-dimensional cross-section along an axis. (d) Qualitative comparison on reconstructions of static machine learning approaches (Baseline (0.5 M) and Baseline (21 M); Baseline (0.5 M) referring to Goy et al, Proc. Natl. Acad. Soc., 116(40), pp. 19848-19856 (2019)) and the dynamical machine learning approach.

Credit

by Iksung Kang, Alexandre Goy, and George Barbastathis

Usage Restrictions

None

License

Licensed content

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.