Prediction of the effectiveness of neoadjuvant chemoradiotherapy in patients with colorectal cancer based on a texture analysis of pretreatment T2-WI magnetic resonance imaging.

Abstract

Backgraund:Recently, significant efforts have been made to search for potential non-invasive biomarkers for predicting the response of locally advanced rectal cancer(LARC) to neoadjuvant chemoradiotherapy(nCRT), among which radiomic features of diagnostic images of the tumor are of the greatest interest, allowing to identify visually indeterminate information about its structure. Aim:To assess the textural features of LARC on pretreatment T2-weighted magnetic resonance images as a potential predictor of the effectiveness of (nCRT) and develop a system for predicting the effectiveness. Materials and methods:The retrospective study included 82 patients with RC were divided into a training sample-58 and 24 – a control sample. A texture analysis(TA) was performed on pretreatment T2-weighted magnetic resonance images. TA performed using the GLCM method and the MaZda ver.4.6. In the training sample, based on morphological assessment, significantly different TA parameters were revealed for two groups: good predict (GP) and poor predict (PP) to treatment based on which a scoring system was created. Results:Groups GP and РP in the training sample differed significantly in texture parameters: AngScMom(р=0,021), SumofSqs(р=0,003), SumEntrp(р=0,003), Entropy(р=0,038) and SumVarnc(р=0,015). It was decided to use 3 parameters to create a scoring system, excluding the Entropy, which had a strong direct correlation with SumEntrp and the lowest AUC, and SumofSqs, which had low reproducibility. The scoring system had asensitivity/specificity/predictive value of a positive result/predictive value of a negative result 72,4%, 69%,7 0%, 71,4% for the training and 80%, 64,3%, 61,6%, 81,8% for the control samples. The area under the ROC curve was 0,77 for the training sample and 0,72 for the control sample. Conclusions:Thus, the texture analysis of pretreatment T2-weighted magnetic resonance images in patients with LARC made it possible to predict the effectiveness of nCRT with good diagnostic efficiency.