Study On Application Value Of MRI Image Texture Analysis For T Staging Of Rectal Cancer

BACKGROUD:Rectal cancer is one of the most common malignancies in the world and one of the common causes of death in cancer patients.Stereotactic radiotherapy,neoadjuvant therapy,molecular targeted therapy and surgical resection are common treatments for rectal cancer.Accurate tumor staging has important implications for the choice of treatment options and the prognosis of patients,and has an important impact on patient prognosis and quality of life.Magnetic resonance imaging(MRI)routine imaging examination is of great value in the diagnosis,evaluation of preoperative staging and prognosis of rectal cancer.Preoperative staging can be performed by observing the location,size,shape,depth of invasion,and relationship with surrounding tissues,and then provide guide the choice of clinical surgical methods.Directly relying on visual observation to make a more accurate judgment of rectal cancer staging is still limited.Texture analysis(TA)is a new image evaluation technology,by analyzing the gray value and pixel value of MR images,we could evaluate tumor staging more effectively and make more accurate and effective personalized therapy.OBJECTIVE: To investigate the texture features of T2 and T3 rectal tumor tissues on MRI images(T2WI),and to evaluate their application value for T staging of rectal cancer.METHODS: MRI images(T2WI)of 81 patients with T2 and T3 rectal cancer who underwent 3.0T magnetic resonance imaging before treatment and biopsy after resection were collected.Image texture was analyzed with MaZda software,and T2 WI texture feature parameters were extracted to screen out the texture feature parameters of the most differential diagnosis value,then the texture parameters were classified and analyzed to assesse the accuracy of differentiating T2 and T3 staging.Three optimal texture feature parameters with the lowest P value were selected by utilizing logisitic regression analysis,and the ROC curve for the selected parameters was plotted,then the area under the curve(AUC)was calculated so as to obtain the optimal diagnostic threshold,specificity and sensitivity.Data analysis was performed using SPSS 22.0 software.RESULTS:(1)General information: of 81 study participants,41 cases are in T2 stage and 40 cases arein T3 stage;there is no statistical difference in the distance from tumor to anal edge and the maximum diameter distribution of tumor between T2 stage and T3 stage patients(P>0.05).(2)Screening and classification analysis methods of T2 WI texture feature parameters: among different screening methods,the accuracy of FPM(62.96%~88.89%)is higher than that of MI(62.96%~85.19%),Fisher(62.96%~85.19%)and POE+ACC(62.96%~85.19%);among classification methods of T2 WI texture features,the accuracy of NDA(85.19% ~ 88.89%)for distinguishing T2 and T3 stages of rectal cancer is higher than that of RDA(65.43% ~ 77.78%),PCA(62.96% ~ 77.78%)and LDA(62.96%~77.78%),in which the highest accuracy of FPM combined with NDA is 88.89%.(3)T2WI texture feature analysis:(1)logistic regression analysis showed that Horzl_GLevNonU,Mean and WavEnLH_s-1 are major influence factors in the T staging of intestinal cancer.Horzl_GLevNonU and Mean were positively correlated with T staging of rectal cancer.The larger the value,the higher the probability of T3 stage in MRI.WavEnLH_s-1 is negatively correlated with T staging of rectal cancer.The smaller the value,the higher the probability of T3 stage in MRI.(2)ROC curve analysis showed that in the case of T staging with Horzl_GlevNonU,AUC = 0.688(95% CI: 0.576-0.787),the cut-off value = 238.974,the sensitivity for T staging as diagnostic reference value = 42.11%,specificity=97.67%,and Youden index = 0.398;in the case of T staging with Mean,AUC=0.788(95% CI:0.683-0.871),cut-off value = 463.876,the sensitivity for T staging as diagnostic reference value = 63.16%,specificity=88.37%,and Youden index = 0.515;in the case of T staging with WavEnLH_s-1,AUC=0.730(95% CI: 0.620-0.823),cut-off value = 59.696,the sensitivity for T staging as diagnostic reference value =81.58%,specificity = 65.12%,and Youden index = 0.467.CONCLUSION:(1)T2WI texture features of rectal cancer in different T stages have statistically significant differences;(2)Horzl_GLevNonU,Mean and WavEnLH_s-1 are major influence factors of T2 and T3 stages of rectal cancer,which are of clinical application value;(3)Conventional MRI T2 WI texture analysis can be used for distinguishing T stages of rectal cancer,and can provide a reliable basis for T-stage diagnosis of rectal cancer.