Can Current Preoperative Imaging Be Used to Detect Microvascular Invasion of Hepatocellular Carcinoma?

To determine the accuracy of imaging features, such as tumor dimension, multinodularity, nonsmooth tumor margins, peritumoral enhancement, and radiogenomic algorithm based on the association between imaging features (internal arteries and hypoattenuating halos) and gene expression that the authors called two-trait predictor of venous invasion (TTPVI), in the prediction of microvascular invasion (MVI) in hepatocellular carcinoma (HCC).

This single-center retrospective study was approved by the institutional review board, and the requirement for informed consent was waived. One hundred twenty-five patients (median age, 63 years; interquartile range, 53-71 years) with a diagnosis of HCC and indications for hepatic resection were included. Two observers independently reviewed radiologic images to evaluate the following features for MVI: maximum diameter, number of lesions, tumor margins, TTPVI, and peritumoral enhancement. Interobserver agreement was checked, and diagnostic accuracy of radiologic features was investigated.

The total number of HCC nodules was 140. Large tumor size, nonsmooth tumor margins, TTPVI, and peritumoral enhancement were significantly related to the presence of MVI (P < .05 in all cases and for both observers). Multinodularity was not significantly related (P = .158). Moreover, the diagnostic accuracy of the three "worrisome" radiologic features (nonsmooth tumor margins, peritumoral enhancement, and TTPVI) was associated with tumor size: The negative predictive value of the absence of worrisome features decreased from 0.84 for observer 1 and 0.91 for observer 2 for tumors smaller than 2 cm to 0.56 and 0.71, respectively, for tumors larger than 5 cm, whereas the presence of all three worrisome features returned to a positive predictive value of 0.95 for observer 1 and 0.96 for observer 2 independent of tumor size, with no significant interobserver differences (P > .10).

"Worrisome" imaging features, such as tumor dimension, nonsmooth tumor margins, peritumoral enhancement, and TTPVI, have high accuracy in the prediction of MVI in HCC.