BACKGROUND: The prevalence and detection rates of adrenal incidentalomas have been on the rise globally, with more than 90% of these lesions pathologically classified as adrenocortical adenomas. Among these, approximately 30% of patients present with hormone-secreting adenomas, leading to the deterioration of their health, with some requiring surgical resection. The available methods for adrenal function evaluation are invasive and costly. Moreover, their accuracy is influenced by numerous factors. Therefore, it is imperative to develop non-invasive and simplified preoperative diagnostic approach. METHODS: A retrospective study was performed on 169 patients from two tertiary medical centers. Subsequently, radiomics features were extracted after tumor margins were delineated layer-by-layer using a semi-automatic contouring approach. Feature selection was achieved in two cycles, with the first round utilizing a support vector machine (SVM) and the second round using a LASSO-based recursive feature elimination algorithm. Finally, logistic regression models were constructed using the clinico-radiological, radiomics, and a combination of both. RESULTS: After a comprehensive evaluation of the predictive indicators, the logistic regression classifier model based on the combined clinico-radiological and radiomic features had an AUC of (0.945, 0.927, 0.856) for aldosterone-producing adenoma (APA), (0.963, 0.889, 0.887) for cortisol-producing adenoma (CPA), and (0.940, 0.765, 0.816) for non-functioning adrenal adenoma (NAA) in the training set, validation set, and external test set, respectively. This model exhibited superior predictive performance in differentiating between the three adrenal adenoma subtypes. CONCLUSIONS: A logistic regression model was constructed using radiomics and clinico-radiological features derived from multi-phase enhanced CT images and conducted external validation. The combined model showed good overall performance, highlighting the feasibility of applying the model for preoperative differentiation and prediction of various types of ACA.