Characterizing International Biomarker Standardization Initiative Image Features using Brodatz Textures.

The objective of this study is to explore the relative effectiveness of International Biomarker Standardization Initiative (IBSI) radiomic metrics in differentiating between the seven diverse texture types of the Brodatz image database. The texture types were established based on a prior study by combining four different classification methods to stratify each of the Brodatz image textures into the seven distinct clusters. Each of these seven texture clusters varied in their contrast, pattern, directionality, uniformity, disorder, and the presence of repeating lines. In this study, 463 IBSI-benchmarked texture metrics spanning eight different texture families, including intensity, histogram, gray level co-occurrence matrix (GLCM), grey level run length matrix (GLRLM), grey level size zone matrix (GLSZM), grey level distance zone matrix (GLDZM), neighborhood grey level dependence matrix (NGLDM), and neighborhood grey tone difference matrix (NGTDM) were extracted from each of the Brodatz images belonging to one of the seven different clusters. Coefficient of variation (COV) was used to assess the distribution of the image features within each cluster and the area under the receiving operating curve (AUC) was used to assess the ability of IBSI-compliant radiomic metrics to differentiate between clusters. We demonstrate that not all IBSI texture metrics assess differences in textures with the same sensitivity.