Application of texture analysis based on apparent diffusion coefficient maps in discriminating different stages of rectal cancer

Abstract

Purpose To explore the potential of texture analysis based on apparent diffusion coefficient (ADC) maps, as a predictor of local invasion depth (stage pT1‐2 versus pT3‐4) and nodal status (pN0 versus pN1‐2) of rectal cancer. Materials and Methods Sixty‐eight patients with rectal cancer underwent preoperative magnetic resonance (MR) imaging including diffusion weighted imaging (DWI) at a 3.0 Tesla system. Routine ADC variables (ADC mean , ADC min , ADC max ), histogram features (skewness, kurtosis) and gray level co‐occurrence matrix features (entropy, contrast, correlation) were compared between pT1‐2 and pT3‐4 stages, between pN0 and pN1‐2 stages. Results Skewness, entropy, and contrast were significantly lower in patients with pT1‐2 as compared to those with pT3‐4 tumors (0.166 versus 0.476, P = 0.015; 3.212 versus 3.441 P = 0.004; 10.773 versus 13.596, P = 0.017). Furthermore, skewness and entropy were identified as independent predictors for extramural invasion of tumors (stage pT3‐4). Significant differences were observed between pN0 and pN1‐2 tumors with respect to ADC mean (1.152 versus 1.044, P = 0.029), ADC max (1.692 versus 1.460, P = 0.006) and entropy (3.299 versus 3.486, P = 0.015). ADC max. and entropy were independent predictors of positive nodal status. Conclusion Texture analysis on ADC maps could provide valuable information in identifying locally advanced rectal cancer. Level of Evidence : 3 J. MAGN. RESON. IMAGING 2017;45:1798–1808