PERSoN4: A multiparametric ultrasound model to improve CEUS LI-RADS for HCC

Background & Aims: Dynamic contrast-enhanced ultrasound (D-CEUS) could be a valuable tool for the non-invasive diagnosis of hepatocellular carcinoma (HCC) with atypical vascular imaging features. Methods: Between January 2021 and November 2023, consecutive patients with chronic liver disease and liver nodules who were candidates for liver biopsy were enrolled in this cohort study. CEUS was performed in all patients before biopsy and categorized according to the CEUS Liver Imaging Reporting and Data System (LI-RADS). Clips were examined using VueBox® software. Clinical and ultrasound parameters were compared among the different histological entities, analyzed with univariable analysis, and incorporated into a logistic regression model for HCC diagnosis. The diagnostic accuracy of the identified model was evaluated by receiver operating characteristic (ROC) curve and relative AUC. The model was then tested on a validation cohort comprising consecutive patients from two centers. Results: A total of 88 patients (57 with HCC, 17 with intrahepatic cholangiocarcinoma, 11 with liver metastases, and three with benign lesions) were enrolled. Statistically significant differences between patients with and without HCC in the training cohort were incorporated in an optimal logistic regression model that included the following predictive variables: sex, number of nodules ≥4, peripheral rim-like hyperenhancement, and peak enhancement (PE) ratio (PE–rim-like enhancement–Sex–Nodules≥4; PERSoN4). The model displayed high accuracy (AUC 0.91) for the diagnosis of HCC. In the validation cohort, the model showed a sensitivity of 48.8% and a specificity of 100.0%, with a positive predictive value (PPV) of 100.0%, maintaining good diagnostic accuracy (AUC of 0.74). Conclusions: PERSoN4 could improve the performance of CEUS LI-RADS criteria, possibly leading to a non-invasive diagnosis of HCC in nearly 50% of patients currently referred for liver biopsy. However, this model requires further external validation before entering clinical practice. Impact and implications: Accurate non-invasive diagnosis of HCC remains challenging in patients with atypical vascular patterns on CEUS, providing the scientific rationale for developing a multiparametric D-CEUS-based risk model that integrates quantitative perfusion analysis with clinical and imaging features. Our findings suggest that the PERSoN4 model could meaningfully enhance the diagnostic performance of CEUS LI-RADS, particularly by identifying a subset of patients in whom HCC can be diagnosed with very high specificity and PPV, which is relevant for hepatologists, radiologists, and multidisciplinary tumor boards managing indeterminate nodules. This approach could reduce the need for liver biopsy in nearly half of currently eligible patients, streamlining diagnostic pathways and potentially lowering procedure-related risks and costs. However, given the moderate sensitivity and the limited sample size, further large-scale external validation is essential before widespread clinical implementation.