The Value Of Predicting Preoperative Blood Supply And Postoperative Recurrence Of Pituitary Macroadenoma Based On MRI Radiomics

PART1The value of predicting preoperative blood supply of pituitary macroadenoma based on MRI radiomicsObjective:To study the value of MRI-based radiomics model in preoperative prediction of the blood supply of pituitary macroadenomaMethods:The clinical and imaging data of 131 patients with pituitary macroadenomas(diameter greater than 10mm)confirmed by pathology in the First Affliated Yijishan Hospital of Wannan Medical College between April,2013 and June,2022 were retrospectively analyzed.According to the intraoperative findings,they were divided into rich blood supply group(45 cases)and general blood supply group(86 cases).All patients were randomly divided into training group(n=93)and validation group(n=38)at a ratio of 7:3.All patients underwent pituitary enhanced magnetic resonance imaging.ITK-SNAP software was used to manually segment the region of interest(ROI)layer by layer on coronal T1WI,T2WI and enhanced T1WI images,respectively,and then 3D fusion was performed.Then AK software was imported to extract texture features of the lesion in each phase.The minimum redundancy and maximum correlation(mRMR)and least absolute value convergence and selection operator(LASSO)were used to screen the texture features for dimensionality reduction and establish the radiomics signature.The score of each patient was calculated,and the reliability of the model was determined by 100 leave-group-out cross validation(LGOCV).Multivariate Logistic regression analysis was used to establish a conventional model including clinical and imaging features with statistically significant differences,and a comprehensive model combined with radiomics model.The receiver operating characteristic(ROC)curve was drawn to evaluate the diagnostic efficacy of the model,and the decision curve analysis(DCA)was drawn to evaluate the net clinical benefit of the model.Results:The AUCs of the T1WI,T2WI,T1WI-enhanced and combined sequence imaging histology models were 0.82,0.78,0.87 and 0.88 in the training group and 0.84,0.79,0.83 and 0.86 in the validation group,respectively.the AUCs of the conventional model and the combined diagnostic model were 0.67 and 0.88 in the training group and 0.82 and 0.86 in the validation group,respectively.the AUCs of the conventional model and the combined diagnostic model were 0.67 and 0.88 in the training group and 0.82 and 0.86 in the validation group,respectively.The efficacy of the conventional model in predicting blood supply of pituitary macroadenoma was poor,and the predictive efficacy of the combined model and T1WI-enhanced serial imaging histology was higher,and DCA showed that the net clinical benefit of the combined model was better than that of the conventional model.Conclusions:The radiomics model has a high value in predicting the blood supply of pituitary macroadenoma,which is better than the judgment of MRI images by clinicians through naked eye observation.It has a good net clinical benefit,and can provide effective guidance for clinical decision-making.PART 2The value of predicting postoperative recurrence of pituitary macroadenoma based on MRI radiomicsObjective:To explore the value of MRI-based radiomics model in predicting the recurrence of pituitary macroadenoma after operation.Methods:Clinical and imaging data of 102 patients with pituitary macroadenoma(diameter greater than 10mm)who were pathologically diagnosed in the First Affiliated Yijishan Hospital of Wannan Medical College from April 2013 to June 2022 were retrospectively collected.After more than one year’s follow-up,all patients were divided into a relapsed group(31 cases)and a non-relapsed group(71 cases)according to the Expert Consensus on Diagnosis and Treatment of recurrent pituitary adenoma in China(2019 edition).According to a completely random method,all patients were divided into the training group(n=72 cases)and the verification group(n=30 cases)in a 7:3 ratio.All patients underwent pituitary enhanced magnetic resonance scanning.ITK-SNAP software was used to delineate the ROI of the lesions layer by layer on the coronal T1WI,T2WI and T1WI enhanced images,and then three-dimensional fusion was performed.The texture features of the lesions at each stage were extracted using AK software.The minimum redundancy maximum correlation(mRMR)and minimum absolute convergence and selection operator(LASSO)were used to screen and reduce the dimension of texture features,and image omics labels were established.The score of each patient was calculated,and the reliability of the model was determined by 100 stay group cross validation(LGOCV).Multivariate Logistic regression analysis was used to establish a conventional model including statistically significant differences in clinical and imaging features,as well as a comprehensive model combined with the image omics model.Receiver operating characteristic(ROC)curves were plotted to evaluate the diagnostic effectiveness of the model,and decision curve analysis(DCA)was plotted to evaluate the clinical net benefit of the model.Results:The AUC of T1WI,T2WI,T1WI enhanced and combined sequence imaging models were 0.83,0.87,0.87,0.89 in the training group and 0.77,0.83,0.79,0.85 in the validation group,respectively.The AUC of the conventional model and the comprehensive diagnosis model in the training group were 0.84 and 0.96,and that in the verification group were 0.86 and 0.87,respectively.The comprehensive model has the highest efficacy in predicting postoperative recurrence of pituitary large adenoma,and DCA shows that the clinical net benefit of the comprehensive model is superior to the conventional model and the imaging omics model.Conclusions:Both the imaging omics model and the integrated model based on clinical and conventional MRI features and texture features have high value in predicting postoperative recurrence of pituitary macroadenoma,and have good clinical net benefits.Imaging omics can provide effective guidance for clinical decision-making.