Radiomic Analysis For Pretreatment Prediction Of Response To Neoadjuvant Chemotherapy In Cervical Cancer

Cervical cancer is the second cancer among women worldwide,and it ranks first in the incidence of gynecological tumors in China.Because the initial symptoms of cervical cancer are not obvious,most patients are locally advanced at the time of diagnosis.The treatment plan for locally advanced cervical cancer is neoadjuvant therapy plus radical resection.For most patients,neoadjuvant therapy can reduce tumor size and improve surgical success.However,some patients still received additional treatment and took risks.Their tumor did not shrink,which delayed the optimal treatment time.Therefore,it is of great clinical significance to predict the response to neoadjuvant therapy before treatment.Based on this clinical problem and current research status,this paper proposes to use radiomics to predict the response to neoadjuvant therapy before treatment.The detailed work is as follows:(1)This article proposes the use of radiomics to predict the response to neoadjuvant therapy in patients with cervical cancer.Multimodality imaging data before neoadjuvant treatment were collected from multiple hospitals.Quantitative radiomic features were extracted from tumor images,and the predictive models were constructed to predict the response to neoadjuvant therapy.This method shown good performance on the data of the test set and could accurately predict the response of patients.(2)To deal with the limits of features extracted from medical images,a combination of image features and clinical information of patients was proposed to construct a more personalized prediction model.By combining clinical information,the prediction accuracy of the model has been improved and passed the statistical test.In addition,the parameters of the radiomic-clinical model were visualized,and a nomogram was drawn for the convenience of the clinician.(3)In order to solve the problem of inadequate representation of the intratumor features,an intra-peritumor radiomics model was proposed by combining intratumoral and peritumoral imaging features.This method makes full use of the differences in tumor microenvironment within and around the tumor,and effectively expands the previous tumor features.The prediction effect of this model is obviously better than the prediction model established by using only intratumoral features.In addition,the robustness of the model was evaluated,which verified the validity of the radiomic features and the robustness of the model.