A study on multiparametric MRI-based radiomics approach to preoperatively estimating tumor-stroma ratio in rectal cancer / 中华放射学杂志

ABSTRACT

Objective:To develop and validate a radiomics biomarker for the preoperative estimation of the tumor-stroma ratio (TSR) in rectal cancer.Methods:From January 2016 to March 2019, totally 149 patients with rectal cancer were enrolled retrospectively at Fudan University Shanghai Cancer Center. The patients were divided into two cohorts using a random number table, 119 in the training and 30 in validation cohorts. The patients were classified into the TSR-high group (TSR>50%) and TSR-low group (TSR≤50%) according to the content of tumor stroma in pathology. All patients underwent T 2WI, enhanced T 1WI and DWI. The lesions on the T 2WI, enhanced T 1WI, DWI and ADC images were delineated and radiomics features were extracted. A radiomics signature (rad-score) was generated by using the least absolute shrinkage and selection operator (LASSO) algorithm. The Spearman correlation coefficients were used to determine the association between rad-score and TSR. The area under the ROC curve (AUC) was used to assess the diagnostic performance of the rad-score. The reliability of the rad-score was quantified by calculating the sensitivity, specificity and accuracy of TSR. Results:With LASSO, a rad-score with 13 radiomics parameters was successfully constructed and was positively correlated with TSR score in the training ( r=0.72, P<0.001) and validation cohorts ( r=0.46, P=0.011). In the training cohort, the AUC of the rad-score was 0.940, with the sensitivity, specificity, accuracy of 100%, 87.3%, 92.4%. In the validation cohort, the AUC was 0.796, with the sensitivity, specificity, accuracy of 83.3%, 67.7%, 73.3%. Conclusions:The rad-score is of promising value for TSR estimation in rectal cancer. It is a promising supplement for patient stratification and may inform decision-making.