| Background:The adrenal gland changes in patients with essential hypertension are hidden,and conventional imaging can only judge its morphology,but it is difficult to find the microscopic changes of adrenal gland.In recent years,the imaging method has been widely used in the diagnosis and differentiation of adrenal hyperplasia and neoplastic lesions,but the correlation between the features of essential hypertension adrenal imaging and common clinical indicators has not been reported.Objective:In this study,standardized radiomics process was applied to obtain the characteristic texture parameters that differentiate the conventional MRI sequences of adrenal gland with normal appearance between patients with essential hypertension and normal control subjects,then these characteristic parameters were applied for spatial labeling,and the RTF-score values of different areas of adrenal gland in patients with essential hypertension were obtained by spatial clustering method.Therefore,the correlation between different adrenal regional heterogeneity and clinical indicators was determined to obtain new biomarkers of adrenal biological image.Materials and methods:In this retrospective study,plain MRI images of the upper abdomen of 2873 patients in the Department of Cardiology,Affiliated Hospital of Yunnan University from January 2018 to October 2022 were collected,and 61 patients with essential hypertension and 44 non-hypertensive subjects were selected as experimental subjects.The texture features of the whole normal adrenal T2 WI and DWI sequence MRI images were extracted.All subjects were randomly divided into training set and test set.After normalization,dimension reduction and feature selection,six machine learning methods,namely support vector machine,linear discriminant analysis,logistic regression,LASSO logistic regression,random forest and decision tree,were used to establish the model.The optimal model was selected by analyzing the receiver operating characteristic curve and calculating the area under ROC curve,and the selected image omics parameters were used to conduct spatial clustering of texture features of adrenal images and calculate RFT-score,which was clustered into high and low groups.Logistic regression was used to analyze the correlation between RFT-score of whole adrenal gland,different parts of cortex and medulla and clinical indicators in patients with essential hypertension.Results:In the test set,the DWI sequence has a good AUC value,and the AUC value,accuracy,specificity and sensitivity of the 16 feature texture parameter model of DWI sequence are 0.825,0.774,0.769 and 0.778 respectively.Furthermore,the DWI sequence parameters of patients with essential hypertension were clustered in adrenal image texture space and the RTF-score scores of different regions were calculated.In the logistic regression multivariate analysis between RTF-score scores and clinical indexes,it was found that the course of hypertension,triglyceride and low density lipoprotein were protective factors for high score of adrenocortical heterogeneity in adrenocortical group.The OR values were 0.87(0.75-0.98,p=0.050),0.32(0.11-0.73,p=0.017)and 0.10(0.01-0.73,p=0.035),respectively.Total cholesterol was a risk factor for high score of heterogeneity in cortical group,with an OR of 10.52(1.64-102.10,p=0.022).In adrenal medulla group,age was a risk factor for high score of adrenal medulla heterogeneity,OR = 1.09(1.03-1.16,p= 0.004).Left ventricular hypertrophy was a protective factor with high score of adrenal medulla heterogeneity,OR = 0.21(0.04-0.86,p= 0.039).Conclusion:The adrenal texture parameters obtained by MRI commonly used sequence can be used to distinguish patients with essential hypertension from normal subjects,and DWI is the most effective.The high heterogeneity score of texture feature spatial clustering of adrenal gland in different anatomical parts of essential hypertension was correlated with clinical indexes. |
PDF Full Text Request