Purpose: Early recurrence (ER) after surgery is related to early death in patients with hepatocellular carcinoma (HCC) after radical resection. To explore the role of preoperative contrast-enhanced ultrasound (CEUS) in predicting ER of HCC after curative resection and to stratify the risk of ER.
Materials and methods: This study evaluated consecutive 556 patients with HCC who were examined by CEUS during the 2 weeks before curative resection between January 2011 and December 2018. ER was defined as intrahepatic and/or extrahepatic recurrence within 2 year after resection of HCC. Univariate and logistic regression analyses were performed to identify independent risk factors for ER after surgical resection of HCC. Recurrence-free time (RFS) rates were analyzed and compared by log-rank test.
Results: ER occurred in 307 (55.2%) of the 556 patients. Univariate and multivariate analyses revealed that a tumor size ≥ 30 mm and satellite nodules seen on CEUS, DL(deep learning) radiomics reoccurrence score based on the frame of image with the maximum intensity of CEUS and an elevated alpha-fetoprotein level were significantly associated with ER (P < .05). Based on the number of predictors present, patients with CEUS LR-5 HCC were stratified into three risk subgroups: risk group 3 (high-risk patients, 4 predictors), risk group 2 (medium-risk patients, 2-3 predictors), and risk group 1 (low-risk patients, 0-1 predictor). The 2-year RFS rate was 19.4% in risk group 3, 40.9% in risk group 2, and 48.1% in risk group 1; the corresponding mean RFS times were 14.0 ± 2.9 months, 43.7 ± 6.6 months, and 55.5 ± 2.8 months, respectively (P < .001).
Conclusions: Tumor size ≥ 30 mm and satellite nodules seen on CEUS, DL radiomics reoccurrence score based on the frame of image with the maximum intensity of CEUS and an elevated alpha-fetoprotein level can predict ER of HCC.
Keywords: Contrast-enhanced ultrasound; Deep learning; Early recurrence; Hepatocellular carcinoma.
© 2024. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.