Establishment of predictive model based on CT radiomics and clinical features and evaluation of its value in predicting efficacy of neoadjuvant chemotherapy in patients with advanced gastric adenocarcinoma
Objective To establish a prediction model based on CT radiomics and clinical features and evaluate its value in predicting the efficacy of neoadjuvant chemotherapy in the patients with advanced gastric adenocarcinoma.Methods A total of 126 patients with advanced gastric adenocarcinoma were treated with neoadjuvant chemotherapy before radical surgery.The efficacy of neoadjuvant chemotherapy was evaluated by tumor regression grading.The patients were divided into training group(88 cases)and testing group(38 cases)according to the ratio of 7∶3.The radiomics features were extracted by sketching the three-dimensional tumor images in the portal vein phase CT enhanced images.The clinical characteristics for the efficacy of neoadjuvant chemotherapy in training group and testing group were analyzed and the clinical characteristics model was established.The efficiency of three models(radiomics model,clinical features model and combined model)for predicting the efficacy of neoadjuvant chemotherapy was evaluated.Results In the training group,neoadjuvant chemotherapy was effective in 44 cases and ineffective in 44 cases.In the testing group,neoadjuvant chemotherapy was effective in 18 cases and ineffective in 20 cases.There was statistical difference in pathological tissue differentiation between the effective group and ineffective group(P＜0.01).Sixteen radiomics features were selected and put into radiomics model.The clinical features model was only pathological tissue differentiation.The predictive efficiency of combined model was better than that of radiomics model and clinical features model in testing group(P＜0.05).Conclusion Based on CT radiomics and clinical features,the models in predicting the efficacy of neoadjuvant chemotherapy for the patients with advanced gastric adenocarcinoma were established,of which the combined model has better predictive efficiency.
NETL
NSTL
万方数据
