Performance Evaluation of Implicit Neural Representations in Diagnostic Fan-Beam CT Imaging
摘要
Recently, implicit neural representation (INR) has been widely applied in computed tomography (CT) reconstruction, achieving impressive results in sparse view reconstruction and metal artifacts reduction with high peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) values. In this chapter, we conduct a comprehensive evaluation of INR’s effectiveness in fan-beam CT applications using the metrics accredited by the American College of Radiology (ACR), such as CT number accuracy, low-contrast detectability, and spatial resolution. Our studies show that, despite demonstrating potential, further refinement of INR-based techniques is needed to fully harness their capabilities in clinical applications, in terms of CT number accuracy, low-contrast detectability, spatial resolution, and free of artifacts.