Quantitative analysis of tumor matrix patterns through statistical and topological texture features

The tumor extracellular matrix has been focused on by newer approaches to cancer therapy owing to its important functions in the process of drug delivery and cellular metastasis. This study aims to characterize tumor extracellular matrix structures in the presence and absence of therapy, as observed on second harmonic generation (SHG) images through both gray-level co-occurrence matrix (GLCM) derived texture features as well as Minkowski Functionals (MF) that focus on the underlying gray-level topology and geometry of the texture patterns. Thirteen GLCM texture features and three MF texture features were extracted from 119 regions of interest (ROI) annotated on SHG images of treated and control samples of tumor extracellular matrix. These texture features were then used in a machine learning task to classify ROIs as belonging to treated or control samples. A fuzzy k-nearest neighbor classifier was optimized using random sub-sampling cross-validation for each texture feature and the classification performance was calculated on an independent test set using the area under the ROC curve (AUC); AUC distributions of different features were compared using a Mann-Whitney U-test. Two GLCM features f3 and f13 exhibited a significantly higher classification performance when compared to other GLCM features (p < 0.05). The MF feature Area exhibited the best classification performance among the MF features while also being comparable to that obtained with the best GLCM features. These results show that both statistical and topological texture features can be used as quantitative measures is evaluating the effects of therapy on the tumor extracellular matrix.