DUG-RECON: A Framework for Direct Image Reconstruction using Convolutional Generative Networks

TL;DR

This paper introduces DUG-RECON, a deep learning framework using convolutional generative networks for direct medical image reconstruction, aiming to replace traditional iterative algorithms with faster, end-to-end solutions.

Contribution

It proposes a novel deep learning architecture with a double U-Net generator for direct sinogram-to-image reconstruction in medical imaging.

Findings

The framework effectively reconstructs images from PET and CT data.

It reduces computational time compared to iterative methods.

The approach demonstrates promising initial results for direct image reconstruction.

Abstract

This paper explores convolutional generative networks as an alternative to iterative reconstruction algorithms in medical image reconstruction. The task of medical image reconstruction involves mapping of projection main data collected from the detector to the image domain. This mapping is done typically through iterative reconstruction algorithms which are time consuming and computationally expensive. Trained deep learning networks provide faster outputs as proven in various tasks across computer vision. In this work we propose a direct reconstruction framework exclusively with deep learning architectures. The proposed framework consists of three segments, namely denoising, reconstruction and super resolution. The denoising and the super resolution segments act as processing steps. The reconstruction segment consists of a novel double U-Net generator (DUG) which learns the sinogram-to-image transformation. This entire network was trained on positron emission tomography (PET) and computed tomography (CT) images. The reconstruction framework approximates two-dimensional (2-D) mapping from projection domain to image domain. The architecture proposed in this proof-of-concept work is a novel approach to direct image reconstruction; further improvement is required to implement it in a clinical setting.