Learned Primal-dual Reconstruction

TL;DR

The paper introduces the Learned Primal-Dual algorithm, a deep learning approach for tomographic reconstruction that unrolls a primal-dual optimization method with neural network-based proximal operators, achieving superior accuracy and efficiency.

Contribution

It presents a novel end-to-end trainable primal-dual neural network for tomographic reconstruction that handles non-linear forward operators and outperforms traditional methods.

Findings

>6dB PSNR improvement on Shepp-Logan phantom

6.6dB PSNR improvement over TV on human phantoms

Only ten forward-back-projection computations needed

Abstract

We propose the Learned Primal-Dual algorithm for tomographic reconstruction. The algorithm accounts for a (possibly non-linear) forward operator in a deep neural network by unrolling a proximal primal-dual optimization method, but where the proximal operators have been replaced with convolutional neural networks. The algorithm is trained end-to-end, working directly from raw measured data and it does not depend on any initial reconstruction such as FBP. We compare performance of the proposed method on low dose CT reconstruction against FBP, TV, and deep learning based post-processing of FBP. For the Shepp-Logan phantom we obtain >6dB PSNR improvement against all compared methods. For human phantoms the corresponding improvement is 6.6dB over TV and 2.2dB over learned post-processing along with a substantial improvement in the SSIM. Finally, our algorithm involves only ten forward-back-projection computations, making the method feasible for time critical clinical applications.