Role of Superb Microvascular Imaging (SMI) vascularity index values and vascularity patterns in the differential diagnosis of malignant liver lesions.

Abstract

Purpose: To evaluate the Superb Microvascular Imaging (SMI) vascular patterns and vascularity index (VI) values of malignant focal liver lesions (FLLs), assess their role in differential diagnosis, and examine interobserver agreement.

Materials and methods: A total of 107 patients (52 males, 55 females; mean age 62 ± 12.8 years, range 25-87) referred to the interventional radiology clinic for FLL biopsy between April 2022 and April 2023 were analyzed. Two radiologists independently assessed the SMI vascular patterns and calculated VI values. Differences among three lesion groups - hepatocellular carcinoma (HCC, n = 16), non-HCC primary liver malignancies (n = 16), and metastases (n = 75) - were evaluated, and interobserver agreement was assessed.

Results: Most metastases (88%) demonstrated hypovascular patterns, while HCCs predominantly exhibited hypervascular patterns (68.7-81.3%). Non-HCC primary malignancies showed no dominant vascular pattern. Significant differences in SMI patterns were observed among lesion types (p = 0.001-0.035). VI values for HCCs (7.53-7.73) were significantly higher than those for non-HCC malignancies (2.73-2.93) and metastases (1.35-1.36) (p = 0.0001). ROC analysis based on VI values yielded AUCs of 0.886-0.887, with a cutoff of 2.92 providing 81.3% sensitivity and 79.1-80.2% specificity for HCC diagnosis. The inter-reader agreement for SMI patterns had a kappa score of 0.634, while the intraclass correlation coefficient (ICC) for VI values was 0.959.

Conclusion: HCCs displayed more hypervascular SMI patterns and significantly higher VI values compared to other malignant FLLs, emphasizing the diagnostic potential of VI in distinguishing HCC from non-HCC tumors. Although metastases primarily exhibited hypovascular patterns and low VI values, no dominant vascular pattern was identified in non-HCC primary liver malignancies. Assessing VI values provided higher interobserver agreement compared to SMI patterns, enhancing objectivity and reproducibility.