MRI texture analysis in prediction of treatment response to neoadjuvant chemoradiotherapy in patients with locally advanced rectal cancer

Abstract

Objective To evaluate the value of MRI texture analysis (TA) in prediction of treatment response neoadjuvant chemoradiotherapy (nCRT) in locally advanced rectal cancer patients. Methods Fifty nine histopathologically-proven rectal adenocarcinoma patients through biopsy treated with nCRT before total mesorectal excision were enrolled in this retrospective study. The first MRI examination (pre-nCRT MRI) was performed before nCRT, and the second one (early nCRT MRI) was performed at the third week of nCRT. The texture parameters values were measured, including mean value, standard deviation, skewness, kurtosis, uniformity, energy, and entropy. Tumoral downstaging was determined by comparing the pre-nCRT clinical T stage(cT stage) with the ypT stage. The patients were divided into downstaging and non downstaging group based on postoperative T staging. Parameters were compared between pre-and early nCRT in terms of averages using Wilcoxon signed-rank test. Downstaging and non downstaging groups were analyzed using Mann-Whitney U test. Multivariate logistic regression analysis was performed using the statistically significant parameters between the two groups as independent variables. ROC analysis was performed on the new independent variables obtained by multi-parameter logistic regression analysis and the single parameter independent variables. The diagnostic efficiency of the parameters were evaluated. Results T-downstaging were found in 28 patients after nCRT. The stdDeviation, kurtosis, and uniformity were significantly different between pre-and early nCRT (P 0.05). The pre-nCRT, uniformity, energy, entropy and the early nCRT mean value, entropy were significantly different in patients with downstaging vs. non downstaging (P<0.05). For the pre-nCRT stdDeviation, uniformity, energy, entropy and the early nCRT mean value, entropy, ROC analysis showed an area under curve(AUC) of 0.69, 0.76, 0.68, 0.67 and 0.65, 0.68, respectively. The multivariate logistic regression analysis for the four pre-nCRT independent variables (stdDeviation, uniformity, energy, entropy) achieved logical variable 1, and the logical variable 1 achieved an AUC of 0.78 to discriminate patients with T-downstaging from patients with non downstaging. The multivariate logistic regression analysis for the two early nCRT independent variables (mean value, entropy) achieved logical variable 2, and the logical variable 2 achieved an AUC of 0.69 to predict T-downstaging. Conclusion Pre-and early nCRT MRI TA in rectal cancer have the efficacy to predict treatment response. Key words: Rectal neoplasms; Magnetic resonance imaging; Treatment outcome; Texture analysis