In a phantom and clinical study involving 24 patients with viable hepatocellular carcinoma (HCC), Graafen et al. evaluated reconstruction kernels for triple-phase photon-counting detector CT (PCD-CT), using 50-keV virtual monoenergetic images. Image quality was assessed quantitatively through contrast-to-noise ratio (CNR) and edge sharpness and qualitatively by three readers.
The softest reconstruction kernels (sharpness level 36) produced the highest CNR across all contrast phases (all p < 0.05) without significantly reducing lesion sharpness. Softer kernels were also rated significantly better for image noise and overall image quality, although image contrast and HCC lesion conspicuity did not significantly differ between kernels. There was also no meaningful image-quality difference between regular body and quantitative kernels at equivalent sharpness levels.
Bottom line: For HCC assessment with photon-counting CT, soft reconstruction kernels provide the best overall image quality. The authors specifically recommend quantitative soft kernels, because they maintain comparable image quality while preserving the possibility of subsequent spectral post-processing.
