Abstract
BACKGROUND: The incidence of adrenal incidentalomas (AIs) is increasing yearly. The early discovery of AIs is helpful to better manage adrenal diseases, especially subclinical primary aldosteronism, Cushing's syndrome and pheochromocytoma. METHODS: In this multicenter retrospective study, a total of 778 patients from three different medical centers were assessed. The two-stage cascade network consisted of a 3D Res-Unet network for adrenal gland segmentation and a classifier for determining the presence of AIs. The segmentation network was mainly evaluated by the Dice similarity coefficient (DSC), and the classifier was evaluated by the area under the receiver operator characteristic curve (AUC), accuracy, sensitivity, and specificity. The Delong test was used to compare the classification performance between the cascade network and manual segmentation. RESULTS: A total of 443 patients were randomly assigned in a 7:3 ratio, stratified sampling, to train and valid sets of the model development cohort, and 335 patients from the three centers were included in the test cohort. In the validation set, the AUC of the model for identifying left AI was 88.15%, and the AUC of the model for identifying right AI was 87.90%. There was no significant difference between model performance and manual segmentation of AIs (p > 0.05). In the test cohort, the cascade network achieved AUC of more than 80% and accuracy of more than 75% for both left and right adrenal glands. CONCLUSIONS: The two-stage cascade network based on a deep learning algorithm can be used for automatic recognition of AIs in nonenhanced CT from different centers.