MRI Radiomics Signature For Predicting The Effectiveness Of Neoadjuvant Therapy For Low And Medium Locally Advanced Rectal Cancer

The first part Multi-factorial Study on Sensitivity of Neoadjuvant Therapy for Preoperative low and Medium Locally Advanced Rectal CancerObjective: With the continuous advancement of diagnosis and treatment technology,preoperative neoadjuvant therapy combined with total mesorectal excision has become the standard treatment strategy for locally advanced rectal cance(LARC).However,due to the heterogeneity of tumors,the efficacy of neoadjuvant therapy for LARC varies widely,and there are many factors that affect the efficacy of preoperative neoadjuvant therapy.This study analyzed the relationship between clinical,blood biochemical examination indicators at baseline and the efficacy of neoadjuvant therapy for low-and middle-stage LARC,and explored the predictive factors for the sensitivity of neoadjuvant radiotherapy before and after LARC.Individualized treatment of locally advanced rectal cancer.Material and Method:A total of 189 patients with low-to-medium-grade LARC who were confirmed by colonoscopy biopsy from May 2013 to September 2016 were retrospectively included.Collect the clinical and blood biochemical examination indicators at baseline: age,gender,tumor diameter,tumor distance from the anal margin,degree of tumor perimeter,peripheral circumcision margin(CRM),tumor wall extravasation(EMVI),TNM staging of tumors.All patients who underwent standard neoadjuvant therapy received TME surgery,and the postoperative tumor shrinkage(TRG)was used as the criterion for judging the sensitivity of neoadjuvant therapy.MRI evaluation:Using Siemens 3.0T superconducting magnetic resonance examination equipment,the scan sequence includes: T1 WI and T2 WI axis,T2 WI sagittal,T1 WI enhanced scan sequence,coronal and sagittal.All patient imaging evaluations were performed by two radiodiagnostic physicians with 3 and 10 years of work experience,respectively.The assessment includes: the degree of tumor pericircle,peripheral circumcision(CRM),tumor wall extravasation(EMVI),and TNM staging of the tumor.Statistical analysis:The measurement data conforms to the normal distribution,which is expressed by x ± s;the measurement data does not conform to the normal distribution,which is expressed by the median;the count data is expressed by examples(%).The relationship between age,gender,tumor distance from the anal margin,tumor invasion of the rectal lumen,serum carcinoembryonic antigen(CEA),CA199 levels,and tumor TNM staging were all analyzed by χ2 test,and then logistic multivariate regression analysis was used to screen for rectal cancer.Independent predictor of neoadjuvant therapy.The data was analyzed using SPSS 22.0 statistical software.Take P = 0.05 as the test standard.Result:After neoadjuvant treatment,TRG 1: 56 cases(29.6),TRG 2: 57 cases(30.1%),TRG 3:38 cases(20.1%),TRG 4: 21 cases(11.2%),TRG 5: 17 cases(9%).The effect was significant after neoadjuvant therapy(TRG1 + TRG2).The univariate analysis results showed that the maximum tumor diameter,CEA,CA199 levels,peripheral circumcision margin(CRM),and extravasational vascular invasion(EMVI)of rectal cancer before neoadjuvant treatment were correlated with neoadjuvant therapy(P <0.05).Multivariate Logistic analysis results showed that tumor maximum diameter,blood carcinoembryonic antigen(CEA),and peripheral circumcision margin(CRM)were independent predictors of the efficacy of neoadjuvant therapy for low-grade LARC.Conclusion:Tumor size,CEA level,and circumcision margin(CRM)status before neoadjuvant therapy are independent predictors for predicting the efficacy of neoadjuvant therapy for LARC,which can guide individualized treatment of advanced rectal cancer and related supplements Treatment;high levels of preoperative CEA,positive peripheral margins,and larger tumors often indicate poor efficacy of neoadjuvant therapy.The second part Locally Advanced Rectal Cancer: Radiomics Signature As a New Biomarker in Preoperative Prediction of Neoadjuvant TherapyObjective: Rectal cancer is a highly heterogeneous tumor.Tumor heterogeneity results in different biological characteristics,so there are also differences in the efficacy of neoadjuvant therapy.With the development of computer technology,radiomics has made it possible to quantify medical image information.It can quantitatively visualize image data to obtain overall tumor information and reflect tumor biological characteristics.The purpose of this study is to establish a model for predicting the efficacy of neoadjuvant therapy for rectal cancer in the middle and low-level locally advanced rectal cancer,and to explore the value of radiomics signature in predicting the efficacy of neoadjuvant therapy in the mid-low local advanced rectal cancer To guide the individualized treatment of locally advanced rectal cancer(LARC)patients.Materials and methods: Retrospectively collected patients with moderate to low-grade locally advanced rectal cancer who were pathologically confirmed from June 2014 to August 2017.All patients underwent TME surgery after neoadjuvant therapy,a total of189 patients.All patients underwent enhanced pelvic MRI scans within 4 weeks before neoadjuvant therapy.The MRI scan sequence is the same as before.Image segmentation: The MRI scan sequences(Axis T1 WI + C,Axis T2 WI,Sag T2WI)of all groups of cases were imported into ITK-sanp software,and the images were segmented by two imaging physicians to obtain 3D regions of interest(ROI).Feature extraction: Based on Matlab2018 a programming,the omics features of segmented rectal cancer lesions were extracted,and a total of 875 imaging omics features were extracted.The intra-group correlation coefficient(ICC)was calculated to evaluate the robustness of the radiomics.The patients were divided into two groups according to the postoperative pathological TRG classification: the neoadjuvant treatment sensitivity group(TRG1-2)and neoadjuvant treatment resistance.Group(TRG3-5)and add imaging omics labels.Dimension reduction is performed by using maximum correlated minimum redundancy(m RMR)and minimum absolute contraction and selection algorithm(Lasso)algorithm.Based on three machine learning classifiers Random Forest(RF),Adjacent K-Space(KNN),Support Vector Machine(SVM),and Fusion Classifier(EC),a prediction model is established for treatment response prediction.Logistic regression was used to construct fitted clinically relevant factors to establish an MRI imaging omics prediction model.Receiver operating characteristic curve(ROC)analysis was drawn to evaluate the diagnostic efficiency of the MRI radiomics prediction model,and decision curves and calibration curves were drawn to evaluate the prediction performance of the radiomics prediction model.Results: The RF model in the machine learning classifier showed a good predictive effect on the efficacy of neoadjuvant therapy for low-to-medium locally advanced rectal cancer.The ROC curve of the verification group(95% CI 0.892-0.997;AUC = 0.830,sensitivity: 0.909,Specificity: 0.879).The 30 radiomics features of the Axis T1 WI C +and Sag T2 WI sequences have shown good predictive ability for the efficacy of neoadjuvant therapy for rectal cancer.An MRI radiomics model that combined CEA and tumor diameter before neoadjuvant therapy in the validation group showed better predictive power: AUC = 0.949(95% CI 0.892-0.997;sensitivity: 0.909,specificity:0.879).Decision curve analysis(DCA)confirmed the clinical value of the nomogram model.Conclusion: The MRI radiomics label can be used as a new biomarker to personalize the prediction of the efficacy of neoadjuvant therapy in patients with locally advanced low-grade rectal cancer before treatment.It can be constructed by combining clinically relevant factors(CEA and tumor diameter).The predictive model’s predictive power can be further improved,which can provide an important basis for the rational formulation of diagnosis and treatment decisions,achieve the maximum benefit for patients,and reduce unnecessary treatment injuries.