Hypervascular hepatic focal lesions on dynamic contrast-enhanced CT: preliminary data from arterial phase scans texture analysis for classification

To investigate the ability of computed tomography texture analysis (CTTA) to distinguish different hypervascular hepatic focal lesions. CTTA software was used to analyse retrospectively 18 cases of focal nodular hyperplasia, 10 cases of hepatic adenoma, 20 cases of haemangioma, 20 cases of hepatocellular carcinoma, and 20 cases of hepatic metastases using arterial phase scans. A list of texture features was generated for lesion classification. Fisher's discriminant analysis (FDA) was used to construct a predictive model from these parameters and to estimate the discriminant accuracy. Receiver operating characteristic (ROC) curve analysis was used to assess the diagnostic performance of texture analysis of benign and malignant tumours. Fifteen texture features were significant differences between the five different histopathological types of all lesions. The total discriminant accuracy was 69.3%, with 55.7% cross-validation accuracy. Seven texture features showed significant differences between the benign and malignant tumours. The total discriminant accuracy in the sample was 83%, with 77.3% cross-validation accuracy. The area under the ROC curve (AUROC) of united texture features was 0.927 (95% confidence interval [CI]=0.875–0.979). CTTA can be used as an aid in the differential diagnosis of hypervascular solid focal hepatic lesions, especially the differential diagnosis between benign and malignant lesions. •Texture analysis is a recently developed image postprocessing technique.•Quantitative changes in texture features can reflect pathological changes in organisms.•Texture analysis provides quantitative measures for tumor lesion discrimination.•Texture analysis of CT images can facilitate differential diagnosis of hypervascular hepatic lesions.