While Model Based Iterative Reconstruction (MBIR) of CT scans has been shown to have better image quality than Filtered Back Projection (FBP), its use has been limited by its high computational cost. More recently, deep convolutional neural networks (CNN) have shown great promise in both denoising and reconstruction applications. In this research, we propose a fast reconstruction algorithm, which we call Deep Learning MBIR (DL-MBIR), for approximating MBIR using a deep residual neural network. The DL-MBIR method is trained to produce reconstructions that approximate true MBIR images using a 16 layer residual convolutional neural network implemented on multiple GPUs using Google Tensorflow. In addition, we propose 2D, 2.5D and 3D variations on the DL-MBIR method and show that the 2.5D method achieves similar quality to the fully 3D method, but with reduced computational cost.
@article{arxiv.1812.08367,
title = {2.5D Deep Learning for CT Image Reconstruction using a Multi-GPU implementation},
author = {Amirkoushyar Ziabari and Dong Hye Ye and Somesh Srivastava and Ken D. Sauer and Jean-Baptiste Thibault and Charles A. Bouman},
journal= {arXiv preprint arXiv:1812.08367},
year = {2018}
}
Comments
IEEE Asilomar conference on signals systems and computers, 2018