| Background: Primary aldosteronism(PA)is a common cause of secondary hypertension.With the popularity of aldosterone-to-renin ratio(ARR)screening,the prevalence of primary aldosteronism rate is higher,nearly 20% in patients with resistant hypertension.Compared with patients with essential hypertension,patients with primary aldosteronism have more cardiovascular and renal outcomes,more higher incidence of metabolic syndrome and diabetes,and less shorter time course of target organ damage.Aldosterone-producing adenoma(APA)and bilateral idiopathic hyperaldosteronism(IHA)are the two common types of primary aldosteronism,accounting for more than 90%.APA can be cured by surgery,while IHA is mainly treated by mineralocorticoid receptor antagonists.Early treatment can reverse the outcome of APA and improve the prognosis.Conventional imaging examination can not identify whether the adrenal nodule is functional or not,so it is impossible to judge whether it is APA.At present,the subtypes of primary aldosteronism are mainly distinguished by adrenal venous sampling(AVS),but AVS has not been carried out in most domestic medical centers.With the development of high-throughput computing,countless quantitative features that can be rapidly extracted from tomographic images,such as CT images.The process of transforming digital medical images into high-dimensional data that can be mined is known as radiomics.Quantitative analyses can reveal a correlation between biomedical images and potential pathophysiology.These data extract quantitative and ideally reproducible information from medical images,including complex patterns that are difficult for the human eye to recognize,which,combined with other relevant data of patients,can be used as complex bioinformatics mining tools to develop models that may improve the accuracy of diagnosis,prognosis and prediction.Objective: The purpose of this study is to use radiomics combined with clinical characteristics to construct a model for the diagnosis and prediction of APA in PA patients with unilateral adrenal adenoma.To provide a non-invasive and feasible scheme for guiding the preoperative diagnosis of APA.Patients and methods: A retrospective study was conducted on 308 patients with primary aldosteronism who underwent AVS in our center from October 2017 to June 2020.According to the inclusion and exclusion criteria,90 PA patients with unilateral adrenal adenoma were included.The whole cohort was divided into training cohort and validation cohort by computer algorithm,and the least absolute shrinkage and selection operator(LASSO)was used to reduce data dimensions and select features.The clinical prediction model,radiomic prediction model and clinical-radiomic prediction model of APA were constructed by multivariable logistic regression.The nomograms were obtained according to their respective models,and their calibration,discrimination and clinical practicability were evaluated.Internal validation was carried out through the validation cohort.Finally,the best prediction model is obtained through the pairwise comparison of Delong’s test,net reclassification index(NRI)and integrated discrimination improvement(IDI).Results: The area under the receiver operating characteristic curve(AUC)of the training cohort of the clinical model was 0.778 [95% confidence interval(CI),0.650 to 0.907],and the AUC of the validation cohort was 0.755 [95% CI 0.572 to 0.938].The specificity,sensitivity,positive predictive value and accuracy of the training cohort and validation cohort were 0.136 and 0.000,0.950 and 1.000,0.667 and 0.750,0.661 and 0.750,respectively.The AUC of the radiomic model training cohort is0.885 [95% CI 0.785 to 0.985],and the AUC of the validation cohort is 0.878 [95%CI 0.701 to 1.000].The specificity,sensitivity,positive predictive value and accuracy of the training cohort and validation cohort are 0.636 and 0.714,1.000 and 0.952,0.833 and 0.909,0.871 and 0.893,respectively.The AUC of the clinical-radiomic model training cohort is 0.900 [95% CI 0.807 to 0.993],and the AUC of the validation cohort is 0.912 [95% CI 0.761 to 1.000].The specificity,sensitivity,positive predictive value and accuracy of the training cohort and validation cohort are0.727 and 0.857,0.975 and 0.905,0.867 and 0.950,0.887 and 0.893,respectively.The training cohort and validation cohort of the three models all passed the calibration(P > 0.05).Delong’s test showed that the AUC of clinical-radiomic model was higher than that of clinical model in the whole cohort and training cohort,and the difference of AUC was statistically significant.The NRI and IDI of the clinical-radiomic model were better than those of the other two models(P < 0.05).Conclusion: A clinical-radiomic model constructed by integrating a radiomic feature and clinical features facilitated accurate prediction of the probability of APA in patients with unilateral adrenal adenoma,and the nomogram constructed on this basis is simple and easy to use,and could be helpful for clinical decision-making,especially for physicians in primary medical hospitals who are unable to carry out AVS. |
PDF Full Text Request