Objectives: To investigate the feasibility of computer-aided discriminative diagnosis among hepatocellular carcinoma (HCC), hepatic metastasis, hepatic hemangioma, hepatic cysts, hepatic adenoma, and hepatic focal nodular hyperplasia, based on radiomics analysis of unenhanced CT images.
Methods: 452 patients with 77 with HCC, 104 with hepatic metastases, 126 with hepatic hemangioma, 99 with hepatic cysts, 24 with FNH, 22 with HA, who underwent CT examination from 2016 to 2018, were included. Radcloud Platform was used to extract radiomics features from manual delineation on unenhanced CT images. Most relevant radiomic features were selected from 1409 via LASSO (least absolute shrinkage and selection operator). The whole dataset was divided into training and testing set with the ratio of 8:2 using computer-generated random numbers. Support Vector Machine (SVM) was used to establish the classifier.
Results: The computer-aided diagnosis model was established based on radiomic features of unenhanced CT images. 27 optimal discriminative features were selected to distinguish the six different histopathological types of all lesions. The classifiers had good diagnostic performance, with the area under curve (AUC) values greater than 0.900 in training and validation groups. The overall accuracy of the training and testing set about differentiating the six different histopathological types of all lesions was 0.88 and 0.76 respectively. 34 optimal discriminative were selected to distinguish the benign and malignant tumors. The overall accuracy in the training and testing set was 0.89and 0.84 respectively.
Conclusions: The computer-aided discriminative diagnosis model based on unenhanced CT images has good clinical potential in distinguishing focal hepatic lesions with noninvasive radiomic features.
Keywords: computer-aided design; diagnosis; liver neoplasms; medical oncology; multidetector computed tomography; radiomics.
Copyright © 2022 Zhao, Liang, Yong, Jia and Gao.