Deep Learning Single View Computed Tomography Guided by FBP Algorithm

Abstract

X-ray Computed Tomography (CT) is widely used in clinical diagnosis. However, the requirement of numerous projections collected in a full-angular range hinders CT image-guided applications such as real-time biopsy. This paper mainly discusses the most challenging single view CT reconstruction problem to speed up the CT-guided clinical workflow. We propose a deep learning approach for single-view CT reconstruction guided by Filtered Back Projection (FBP) algorithm which makes the single view reconstruction accurate, fast and interpretable. We formulate an end-to-end framework that contains the projection generation network to predict sufficient projections from a single view, the FBP layer to obtain coarse CT volume, and the CT fine-tuning network to output the final CT volume. We carefully design our training strategy to ensure the network towards CT reconstruction. Our experiments on the public 4D CT datasets prove that our method achieves state-of-the-art performance.