"Discriminating Statistical Textural Features for the Detection of Pneumonia from X-Ray images"

"This dataset presents a comprehensive radiomic feature knowledge base constructed from chest X-Ray (CXR) images of the publicly available Kermany dataset. The primary goal is to enable the identification of heterogeneous tissue textures associated with different pneumonia variants\u2014viral, bacterial, and healthy lungs\u2014through extensive feature engineering. Both global and local statistical texture features are extracted from segmented lung regions using advanced preprocessing, including lung ROI segmentation. The 846 radiometric features encompass first-order statistics, shape descriptors, and multi-scale transformations such as wavelet, exponential, logarithmic, square, and square root filters. In addition, local texture descriptors like Gray Level Co-occurrence Matrix (GLCM), Gray Level Run Length Matrix (GLRLM), Gray Level Size Zone Matrix (GLSZM), Gray Level Dependence Matrix (GLDM), Laplacian of Gaussian (LoG), Gradient Magnitude, Local Binary Patterns (LBP), and Neighborhood Gray-Tone Difference Matrix (NGTDM) are computed. Feature dimensionality is reduced based on variance explanation capacity to form a discriminative knowledge base. This curated feature repository supports non-linear classification tasks and facilitates pneumonia variant differentiation through quantifiable radiomic biomarkers."