Locally advanced rectal cancer: predicting non-responders to neoadjuvant chemoradiotherapy using apparent diffusion coefficient textures.

Abstract

The purpose of the study is to evaluate whether apparent diffusion coefficient (ADC) textures could identify patient with locally advanced rectal cancer (LARC) who would not respond to neoadjuvant chemoradiotherapy (NCRT). Twenty-six patients who underwent MRI including diffusion-weighted imaging at a 3.0T system before NCRT were enrolled. Texture analysis of pre-therapy ADC mapping was carried out, and a total of 133 ADC textures as well as routine mean ADC value of the primary tumor were extracted for each patient. Texture parameters and mean ADC were compared between responsive group and non-responsive group. Logistic regression was used to determine the independent predictors for non-responders. Receiver operating characteristic curve (ROC) was performed to evaluate the predictive performance of the significant parameters. Eighteen of the 133 texture parameters significantly differed between responsive and non-responsive groups (p<0.05). Further, energy variance and SdGa47 were identified as independent predictors for non-responders to NCRT; this logistic model achieved an area under the curve (AUC) of 0.908. Texture analysis based on pre-therapy ADC mapping could potentially be helpful to identify patients with LARC who would not respond to NCRT.