The value of CT-based radiomics models in differentiating adrenal adenoma subtypes
WU Chunmei
LI Xuanang
WANG Weicheng
XUE Linyan
YIN Xiaoping
Abstract：Objective To investigate the clinical value of multiphase CT radiomics models in differentiating adrenal adenoma subtypes.Methods A retrospective collection of 195 patients with pathologically confirmed adrenal adenomas who underwent preoperative non-contrast and contrast-enhanced CT scans was conducted.Patients were randomly divided into a training set(156 cases)and a validation set(39 cases)in an 8∶2 ratio.Based on the hormone secretion level,the adrenal adenomas were divided into functional adrenal adenoma group(70 cases)and non-functional adrenal adenoma group(125 cases).Based on hormone secretion levels,patients were divided into a functional adrenal adenoma group(70 cases)and a non-functional adrenal adenoma group(125 cases).A total of 1521 radiomics features were extracted from the patients'non-contrast and contrast-enhanced CT images.Using the random forest algorithm,the top 8 optimal radiomics features were selected.Radiomics models were constructed using the support vector machine(SVM)algorithm with 5-fold cross-validation for the non-contrast phase,arterial phase,venous phase,delayed phase,and a model combining all phases.Clinical and CT imaging features were compared between the two groups using t-tests,Mann-Whitney U tests,and chi-square tests.The receiver operating characteristic(ROC)curve was plotted,and the area under the curve(AUC),accuracy,sensitivity,and specificity were calculated to evaluate the model's performance.Results The functional adrenal adenoma group had lower blood potassium levels and a higher proportion of patients with hypertension compared to the non-functional adrenal adenoma group(both P＜0.05).In the validation set,the non-contrast phase radiomics model showed the highest diagnostic performance with an AUC of 0.813,accuracy of 0.759,and specificity of 0.814.Conclusion The radiomics model based on CT radiomics features from different phases combined with SVM can non-invasively differentiate adrenal adenoma subtypes.This model has significant clinical value and can aid in clinical decision-making.
Keywords：Adrenal adenomaRadiomicsMachine learningSupport vector machinesRandom forestsTomographyX-ray computed
Publication Date：2024-05-15
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：7( 274-279,287 )