Radiomics For Predicting Radiation Enteritis In Cervical Cancer

BackgroundCervical cancer is the fourth most common cancer among female malignancies in developed countries behind breast,colorectal and lung cancers,which 87% of cases occur in developing countries,and it is also the leading cause of death.Different treatment modalities are chosen for patients with cervical cancer depending on the stage and lymph node metastasis.Early stage cervical cancer patients are often treated with surgery,while simultaneous radiotherapy is often recommended for locally advanced patients.As one of the main tools in the radical treatment of cervical cancer,radiotherapy plays an important role.However,toxic reactions caused by radiation therapy are difficult to predict and avoid.In the context of precision medicine,finding a method that can distinguish between high and low-risk patients at an early stage,effectively guiding individualized treatment and improving the quality of patient survival is one of the urgent problems we need to address in our clinical treatment work at present.Imaging histology can extract features from somatic images in a rapid and large number,further analyze high-dimensional data,quantify the features,and by quantifying high-dimensional data and establishing specific imaging histology labels,predictive models can be developed.This study will also provide a comparative discussion of the better models for handling high-dimensional data.ObjectivesExplore omics tags based on CT image in cervical cancer after chemoradiation radioactive enteritis predictive value,and develop clinical is easy to use tool,to help clinicians better found in the clinical diagnosis and radioactive enteritis of high-risk patients after radiotherapy,and to prevent in advance,to further improve survival in patients with individualized precision medical quality and development.MethodsIn a study to investigate the prediction of CT-based imaging histology labeling in the development of radiation enterocolitis after concurrent radiotherapy for cervical cancer,227 patients with cervical cancer who received concurrent radiotherapy were retrospectively included.These patients had a pathologically confirmed diagnosis of squamous cell carcinoma at the initial diagnosis and were followed regularly after the completion of radiotherapy.Relevant clinicopathological factors,basic imaging features were analyzed,texture features were screened from pre-radiotherapy CT images by LASSO regression models and imaging histology labels were established,and SVM,random forest and logistic prediction models were established and compared based on the occurrence of radiation enteritis and chronic radiation enteritis within 2 years and 4years after the end of radiotherapy.The subject operating characteristic curves（ROC）of the models with better prediction of radiation enteritis were drawn,and the area under the curve（AUC）and model accuracy were calculated separately.The optimal model was selected to calculate the imaging histology score for each patient and to assess the ability of the imaging histology score in distinguishing between patients with high and low risk.ResultGeneral data:（1）A total of 227 medical records met the inclusion criteria in this study,including 137（60.35%）stage II patients and 90（39.65%）stage III patients;（2）Age distribution ranged from 34 to 75 years old,with an average of 54.6years old.（3）Severe acute radiation enteritis occurred in 10 cases（7.30%）among stage II patients;chronic radiation enteritis occurred in secondary and above radiation enteritis in 31 cases（22.63%）,of which 27 cases（19.71%）occurred within 2 years;acute radiation enteritis in 10 cases（11.11%）and chronic radiation enteritis in 28cases（31.11%）in stage III patients and above,and 23 cases（25.56%）occurred within 2 years.Clinical univariate analysis: there was a statistically significant difference between pre-radiotherapy anemia in the occurrence of radiation enteritis after radiotherapy 2years（P<0.01）.Staging,age,and menopausal status were not statistically significant.Radiomic label: among three prediction models,the Logistic prediction model performed better（accuracy 0.735,AUC0.741）;in the establishment of the prediction model for chronic radiation enteritis within 4 years,the Logistic prediction model also performed better（accuracy 0.748）.Patients were successfully classified into high-risk and low-risk groups（P < 0.001）based on the imagomics scores which derived from the predictive model.ConclusionRadiomic label is an independent predictor of radiation enteritis after concurrent chemoradiotherapy for cervical cancer.In the case that clinical indicators can not predict the occurrence of complications,the existence of imaging omics label plays an important role in predicting the occurrence of radiation enteritis within 2 years after radiotherapy.The logistic regression model which based on imagomics labels has better prediction performance.