Abstract
Deep Learning is developing interesting tools that are of great interest for inverse imaging applications. In this work, we consider a medical imaging reconstruction task from subsampled measurements, which is an active research field where Convolutional Neural Networks have already revealed their great potential. However, the commonly used architectures are very deep and, hence, prone to overfitting and unfeasible for clinical usages. Inspired by the ideas of the green AI literature, we propose a shallow neural network to perform efficient Learned Post-Processing on images roughly reconstructed by the filtered backprojection algorithm. The results show that the proposed inexpensive network computes images of comparable (or even higher) quality in about one-fourth of time and is more robust than the widely used and very deep ResUNet for tomographic reconstructions from sparse-view protocols.
Highlights
Convolutional Neural Networks (CNNs), with their remarkable capacity of learning with multiple levels of abstraction, are giving new impetus to researchers working on inverse problems, and the imaging sector is one of the most involved field [1]
To simulate the sparse-view geometry, we considered two different protocols: a full angular acquisition with 1-degree spaced projections and a reduced scanning trajectory limited to 180 degrees with 180 projections
We first remark the very poor values achieved by the Filtered Back-Projection (FBP), which are motivated by its difficulty in recovering the actual intensities of the ground truth images
Summary
Convolutional Neural Networks (CNNs), with their remarkable capacity of learning with multiple levels of abstraction, are giving new impetus to researchers working on inverse problems, and the imaging sector is one of the most involved field [1]. Researchers have begun to tackle inverse imaging applications, such as denoising, deconvolution, in-painting, superresolution, and medical image reconstruction, with CNNs, and they all report significant improvements over state-of-the-art techniques, encompassing sparsity-based models derived from compressed sensing approaches [2,3]. Due to limitations of human anatomy or equipment manufactory, in special cases, the X-ray source may walk only a semi-circular or C-shape path (as depicted in Figure 1c), and the SpCT configuration is labeled as limited angle CT. Such low-dose tomographic approaches lead to incomplete CT projection data, and such subsampled measurements usually produce severe streaking artefacts on the Filtered Back-Projection (FBP) reconstructions. The very accurate achievable results, the optimization approach has not been widely adopted yet in clinical setting because of its high computational cost
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
