Texture analysis of CT images based on machine learning to identify whether adrenal cortical adenoma is functional or not

Background: The detection rates and surgical resection rates of adrenal incidentaloma are increasing, but malignant tumors only account for 10%-15%. Taking adrenal cortical adenomas (ACA) as an example, this paper aims to explore if texture analysis of preoperative computed tomography (CT) images based on machine learning can reliably identify if ACA are functional. Methods: The clinical and imaging data were collected retrospectively for 75 patients with adrenal cortical adenoma confirmed by surgery and pathology in our hospital from November 2018 to November 2020. MaZda image analysis software was used to segment image data and extract features. Support vector machine (SVM) was used to create an image omics model to predict the functionality of the ACA. Lastly, the area under the receiver operating characteristic (ROC) curve (AUC) was used to evaluate the performance of the radiohistology model. Results: Four histological models were developed (models 1-4, which represent plain scan, arterial enhancement, venous enhancement and delayed scan, respectively). All models 1-4 showed a good ability to distinguish between functional and non-functional ACA in the training sample, with an average AUC of 0.96, 0.91, 0.91 and 0.88, respectively. Conclusion: Texture analysis of CT images using machine learning can effectively identify whether adrenal cortical adenoma is functional.