Prediction of Clinical Pathologic Prognostic Factors for Rectal Adenocarcinoma: Volumetric Texture Analysis Based on Apparent Diffusion Coefficient Maps

Texture analysis has been used to characterize and measure tissue heterogeneity in medical images. The purpose of this study was to investigate the potential of texture features derived from apparent diffusion coefficient (ADC) maps, to serve as imaging markers for predicting important histopathologic prognostic factors in rectal cancer. One hundred patients of rectal cancer received 3 T preoperative magnetic resonance imaging including diffusion-weighted imaging (DWI). Skewness, kurtosis, uniformity from the histogram and entropy, energy, inertia, correlation from gray-level co-occurrence matrix (GLCM) derived from whole-lesion volumes were measured. Independent sample t-test or Mann-Whitney U-test and receiver operating characteristic (ROC) curves were used for statistical analysis. Uniformity, energy and entropy were significantly different (p = 0.026, 0.001, and 0.006, respectively) between stage pT1-2 and pT3-4 tumors. Skewness, kurtosis and correlation were significantly different (p = 0.000, 0.006, and 0.041, respectively) between grade 1-2 and grade 3 tumors. Energy and entropy (p = 0.008 and 0.033, respectively) could significantly differentiate negative circumferential resection margin (CRM) from positive CRM. Furthermore, predicted probabilities derived by logistic regression analysis yielded greater area under the curve (AUC) in differentiating pT3-4 stage and grade 3 grade tumors. Texture features derived from ADC maps may useful to predict important histopathologic prognostic factors of rectal cancer.