Noise-resilient deep tomographic imaging

Published

April 24, 2023

Author(s)

Zhen Guo, Zhiguang Liu, George Barbastathis, Qihang Zhang, Michael Glinsky, Bradley Alpert, Zachary H. Levine

Abstract

X-ray tomography is a non-destructive imaging technique that reveals the interior of an object from its projections at different angles. Under limited-angle and low-photon sampling, a regularization prior is required to retrieve a high-fidelity reconstruction. Recently, deep learning has been used in X-ray tomography. The prior learned from training data replaces the general-purpose priors in iterative algorithms, achieving high-quality reconstructions with a neural network. Previous studies typically assume the noise statistics of testing data is acquired \textita priori} from training data, leaving the network susceptible to a change in the noise characteristics under practical imaging conditions. In this work, we propose a noise-resilient deep-reconstruction algorithm for X-ray tomography. By utilizing total variation regularization to prepare the input reconstructions for the network, the learned prior shows strong noise resilience without training with noisy examples. The advantages of our framework may further enable low-photon tomographic imaging where long acquisition times limit the ability to acquire a large training set.

Citation

Optics Express

Volume

31

Issue

10

Pub Type

Journals

Download Paper

https://doi.org/10.1364/OE.486213

Local Download

Keywords

neural network, deep learning, regularization, sparsity, integrated circuit

Citation

Guo, Z. , Liu, Z. , Barbastathis, G. , Zhang, Q. , Glinsky, M. , Alpert, B. and Levine, Z. (2023), Noise-resilient deep tomographic imaging, Optics Express, [online], https://doi.org/10.1364/OE.486213, https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=936008 (Accessed April 27, 2024)