Evaluation Of The Efficacy Of Neoadjuvant Chemoradiotherapy For Rectal Adenocarcinoma Based On Multimodal MRI Radiomics

Objective:To investigate the value of multimodal MRI-based radiomics in evaluating the efficacy of neoadjuvant chemoradiotherapy for rectal adenocarcinoma,and looking for radiomics features associated with efficacy to improve the accuracy of evaluating the efficacy of neoadjuvant chemoradiotherapy for rectal adenocarcinoma,a nomogram provides a reference for clinical selection of personalized treatment methods.Methods:The clinical and imaging data of 70 patients diagnosed with rectal cancer who underwent total mesorectal excision after neoadjuvant chemoradiotherapy from August 2017 to October 2021 in the Second Affiliated Hospital of Nanchang University were retrospectively analyzed.All patients underwent pelvic MRI plain scan and enhanced examination,and postoperative pathology all had complete tumor regression grading.According to AJCC tumor regression grading standards,the patients were divided into a good response group（AJCC grade 0-1）and a poor response group（AJCC grade 2-3）.The volume of interest of the lesion was delineated layer by layer on the axial T₂WI,DWI and enhanced arterial phase LAVA sequence images of each patient,and the dimensionality reduction was performed after the radiomic features were extracted,and all patients were divided into a train group and a test group in a ratio of 7:3.Finally,logistic regression models Model 1,2,3 based on axial T₂WI,DWI,arterial enhancement phase and a joint model 4 based on radiomics signatures are established.Drawing receiver operating characteristic curve and clinical decision curve to evaluate model performance,and drawing a nomogram for individualized evaluation of NCRT efficacy in patients with rectal adenocarcinoma,and SPSS was used for statistical analysis,α=0.05 was taked as the inspection level.Results:Pathological grading of 70 cases rectal adenocarcinoma after neoadjuvant chemoradiotherapy,there were 26 cases in the good curative effect group and 44cases in the poor curative effect group.There was no statistical difference in gender,age,drug use and tumor index between the two groups（P>0.05）.106 features were extracted from each sequence of images of each patient.After feature dimension reduction,there were 4,4 and 9 features left in the T₂WI,DWI and enhanced arterial phase LAVA sequences.Spearman correlation analysis was performed on these features,and it was found that Gray Level Non Uniformity Normalized in T₂WI images,Cluster Shade in DWI images,and entropy in enhanced arterial phase images of tumor volume of interest are negatively correlated with the efficacy of NCRT in rectal adenocarcinoma（P<0.05）.The AUC values of Model 1～4 were 0.838,0.718,0.723 and 0.862 respectively,delong test showed there were statistical difference between Model 4 and Model 2 and 3（P<0.05）,and there was no statistical difference between the other（P>0.05）.DCA showed that all four models could benefit clinically and Model 4 benefited the most.Conclusion:Multimodal MRI-based radiomics has a certain value in evaluating the efficacy of neoadjuvant chemoradiotherapy for rectal adenocarcinoma,Gray Level Non Uniformity Normalized in T₂WI images,Cluster Shade in DWI images,and entropy in enhanced arterial phase images of tumor volume of interest may be radiomics markers reflecting poor NCRT response,and the drawn joint model nomogram has a certain reference value for clinical personalized treatment.