Effects of Reconstruction Filter Kernels in Computed Tomography on Radiomic Image Features

Radiomic metrics offer a promising quantitative tool for diagnosis. This paper seeks to assess the robustness of International Biomarker Standardization Initiative (IBSI) image features across 58 reconstruction filter kernels employed by the Canon Aquilion One CT scanner. By using volumes generated from the same raw data but different image filter kernels, the radiomic metrics generated solely reflect the differences in scanner image reconstruction. Explored are 5 different tissue regions of interest with 463 IBSI image features. Quantification of the image kernel differences could serve future harmonization algorithms.