MARGANVAC: metal artifact reduction method based on generative adversarial network with variable constraints

Abstract

Objective. Metal artifact reduction (MAR) has been a key issue in CT imaging. Recently, MAR methods based on deep learning have achieved promising results. However, when deploying deep learning-based MAR in real-world clinical scenarios, two prominent challenges arise. One limitation is the lack of paired training data in real applications, which limits the practicality of supervised methods. Another limitation is that image-domain methods suitable for more application scenarios are inadequate in performance while end-to-end approaches with better performance are only applicable to fan-beam CT due to large memory consumption. Approach. We propose a novel image-domain MAR method based on the generative adversarial network with variable constraints (MARGANVAC) to improve MAR performance. The proposed variable constraint is a kind of time-varying cost function that can relax the fidelity constraint at the beginning and gradually strengthen the fidelity constraint as the training progresses. To better deploy our image-domain supervised method into practical scenarios, we develop a transfer method to mimic the real metal artifacts by first extracting the real metal traces and then adding them to artifact-free images to generate paired training data. Main results. The effectiveness of the proposed method is validated in simulated fan-beam experiments and real cone-beam experiments. All quantitative and qualitative results demonstrate that the proposed method achieves superior performance compared with the competing methods. Significance. The MARGANVAC model proposed in this paper is an image-domain model that can be conveniently applied to various scenarios such as fan beam and cone beam CT. At the same time, its performance is on par with the cutting-edge dual-domain MAR approaches. In addition, the metal artifact transfer method proposed in this paper can easily generate paired data with real artifact features, which can be better used for model training in real scenarios.

Nonconvex weighted variational metal artifacts removal via convergent primal-dual algorithms

“…With the remarkable success of deep learning in medical image processing, recent studies have implemented deep neural networks (DNNs) to tackle the problem of reducing metal artifacts. The existing researches consist of the image-to-image learning [15][16][17][18][19], the sinogram domain network [20][21][22][23][24] and dual domain (both the image and sinogram domains) [25][26][27][28][29] that utilizes either residual learning or adversarial learning techniques. However, DNNs commonly require large, representative training datasets and extensive computational resources, imposing practical limitations on their widespread adoption.…”

HEAL: High-Frequency Enhanced and Attention-Guided Learning Network for Sparse-View CT Reconstruction

“…In recent years, deep learning has been widely applied in the field of CT imaging and has achieved some encouraging results [10][11][12][13], especially in sparse-view reconstruction, showing better imaging results than compressive sensing models [14,15]. Currently, deep learning-based sparse-view reconstruction methods can be categorized into four types: single-domain learning, direct mapping between measurement data and reconstructed images using networks, network models based on iterative reconstruction algorithms and dual-domain learning.…”