FREQUENCY RESPONSE AND DISTORTION PROPERTIES OF RECONSTRUCTION ALGORITHMS IN COMPUTED TOMOGRAPHY

CT image denoising methods for image quality improvement and radiation dose reduction

With the ever-increasing use of computed tomography (CT), concerns about its radiation dose have become a significant public issue. To address the need for radiation dose reduction, CT denoising methods have been widely investigated and applied in low-dose CT images. Numerous noise reduction algorithms have emerged, such as iterative reconstruction and most recently, deep learning (DL)-based approaches. Given the rapid advancements in Artificial Intelligence techniques, we recognize the need for a comprehensive review that emphasizes the most recently developed methods. Hence, we have performed a thorough analysis of existing literature to provide such a review. Beyond directly comparing the performance, we focus on pivotal aspects, including model training, validation, testing, generalizability, vulnerability, and evaluation methods. This review is expected to raise awareness of the various facets involved in CT image denoising and the specific challenges in developing DL-based models.

Visualization of the distortion induced by nonlinear noise reduction in computed tomography

Purpose

We developed a method to visualize the image distortion induced by nonlinear noise reduction algorithms in computed tomography (CT) systems.

Approach

Nonlinear distortion was defined as the induced residual when testing a reconstruction algorithm by the criteria for a linear system. Two types of images were developed: a nonlinear distortion of an object (NLD object ) image and a nonlinear distortion of noise (NLD noise ) image to visualize the nonlinear distortion induced by an algorithm. Calculation of the images requires access to the sinogram data, which is seldomly fully provided. Hence, an approximation of the NLD object image was estimated. Using simulated CT acquisitions, four noise levels were added onto forward projected sinograms of a typical CT image; these were noise reduced using a median filter with the simultaneous iterative reconstruction technique or a total variation filter with the conjugate gradient least-squares algorithm. The linear reconstruction technique filtered back-projection was also analyzed for comparison.

Results

Structures in the NLD object image indicated contrast and resolution reduction of the nonlinear denoising. Although the approximated NLD object image represented the original NLD object image well, it had a higher random uncertainty. The NLD noise image for the median filter indicated both stochastic variations and structures reminding of the object while for the total variation filter only stochastic variations were indicated.

Conclusions

The developed images visualize nonlinear distortions of denoising algorithms. The object may be distorted by the noise and vice versa. Analyzing the distortion correlated to the object is more critical than analyzing a distortion of stochastic variations. The absence of nonlinear distortion may measure the robustness of the denoising algorithm.