A Bayesian Nonparametric Model for Textural Pattern Heterogeneity

Abstract Cancer radiomics is an emerging discipline promising to elucidate lesion phenotypes and tumour heterogeneity through patterns of enhancement, texture, morphology shape. The prevailing technique for image texture analysis relies on the construction synthesis grey-level co-occurrence matrices (GLCM). Practice currently reduces structured count data a GLCM reductive redundant summary statistics which requires variable selection multiple comparisons each application, thus limiting reproducibility. In this article, we develop Bayesian multivariate probabilistic framework unsupervised clustering sample objects. By appropriately accounting skewness zero inflation observed counts simultaneously adjusting existing spatial autocorrelation at nearby cells, methodology facilitates estimation pattern distributions within lattice itself. techniques are applied cluster images adrenal lesions obtained from CT scans with without administration contrast. We further assess whether resultant subtypes clinically oriented by investigating their correspondence pathological diagnoses. Additionally, compare performance class machine learning approaches used in cancer simulation studies.