Computed tomography ring artifact correction method with super-pixel segmentation and adaptive relative total variation.

Background: Ring artifacts are a common and persistent issue in computed tomography (CT) imaging, arising from detector calibration errors, gain variations, and other hardware inconsistencies, and their presence can significantly compromise the diagnostic accuracy and clinical utility of CT scans. This study aimed to propose a comprehensive approach for removing the ring artifacts in CT images, with a particular focus on the challenges posed by the cutting-edge photon counting CT (PCCT) technology. Methods: This method skillfully combines super-pixel segmentation with an adaptive form of Relative Total Variation, resulting in a robust, two-stage process for artifact correction. In the first stage, high-intensity ring artifacts are efficiently addressed using super-pixel segmentation. The second stage involves the application of adaptive relative total variance, fine-tuned by the mesh adaptive direct search algorithm, to effectively correct low-intensity artifacts. This dual-stage strategy is key to preserving crucial image details while successfully eliminating artifacts. The effectiveness of the proposed method is thoroughly validated through various experiments, including digital simulations and studies involving phantoms and patients. Results: These tests reveal a significant reduction in both high and low-intensity ring artifacts, while maintaining the structural integrity and grayscale balance of the images. Additionally, the versatility and robustness of this method are highlighted by its suitability for a wide range of imaging scenarios and equipment. Conclusions: This study not only tackles a significant challenge in medical imaging but also paves the way for new applications of PCCT in precision medicine. The code is publicly available for reproducibility, Github: https://github.com/XiaoNa-gdmu/Ring.