Clinical Significance Of Pretreatment CT Radiomic-based Nomogram For Induction Chemotherapy In Locally Advanced Hypopharyngeal Cancer

ObjectivePurpose:We aimed to use a new imaging data extraction method,radiomic to predict the efficacy of induction chemotherapy in patients with locally advanced hypopharyngeal cancer before treatment.Materials and MethodsMethod:One hundred and fifty three patients who were newly diagnosed as hypopharyngeal cancer and received induction chemotherapy were included(primary cohort:n=77;validation cohort:n=76).Radiomic features were extracted in each volume of interest of non-contrast CT(NCCT)and contrast-enhanced CT(CECT).Of these,the minimum redundancy maximum relevance(MRMR)algorithm was used to weight and select imaging features.Support vector machine support vector machine(SVM),random forest(RF)and artificial neural network(ANN)classification algorithms were compared through internal validation(100 times 10-fold cross-validation)in the primary cohort,and then the model with the best performance was chosen.Clinical characteristics were analyzed by multivariate analysis and the clinical model was selected by minimum Akaike’s information criterion(AIC).The performance of an overall nomogram integrated the radiomic signatures and remarkable clinical data was evaluated with receiver operating characteristics curves(ROC).ResultsResult:The radiomic signatures were built by SVM from NCCT and CECT.SVM-NCCT and SVM-CECT were significant in primary cohort and validation cohort for predicting treatment response of induction chemotherapy(AUC:0.8920.838).Meanwhile,age and short axis of maximum enlarged lymph node(LN size)were selected among the clinical data.Radiomic signatures model(R=-4.2599 +3.3975xSVM-NCCT + 2.6872×SVM-CECT)and clinical model(C=-0.4559-1.4728×Age+0.3493xLN size)were integrated to develop a nomogram.At last,the overall nomogram has shown the best performance in primary and validation cohort with good calibration(AUC:0.913 0.864).Decision curve analysis had also proved the clinical value of this comprehensive nomogram.ConclusionsConclusion:Pretreatment CT-based radiomic nomogram provided an innovative method to predict the response of induction chemotherapy in patients with locally advanced hypopharyngeal cancer,which help select non-response patients for surgery or concurrent systemic chemoradiotherapy before giving induction chemotherapy.