基于RFA后钆塞酸二钠增强MRI消融内及消融周围影像组学预测肝细胞癌短期预后

OBJECTIVE: To evaluate the predictive value of intra-ablational and peri-ablational radiomics features extracted from the first post-RFA follow-up gadoxetic acid disodium (Gd-EOB-DTPA)-enhanced MRI for short-term poor prognosis after radiofrequency ablation (RFA) in hepatocellular carcinoma (HCC) patients, particularly for local tumor progression.

METHODS: A total of 181 HCC patients who underwent ultrasound-guided RFA at Lanzhou University First Hospital between 2017 and 2025 were retrospectively enrolled. Gd-EOB-DTPA-enhanced MRI and clinical data from the first post-RFA follow-up were collected. Short-term poor prognosis was defined as the appearance of new lesions within 6 months after RFA. Radiomics models based on MRI (arterial phase, portal venous phase, hepatobiliary phase) across various regions (entire ablation zone, Peri 3 mm, 5 mm, 7 mm) were developed using the LASSO approach and logistic regression. A combined model was constructed using best radiomics model features and clinical-radiological predictors. The performance, calibration and clinical utility were evaluated by ROC curve, calibration curve, and decision curve analysis, respectively.

RESULTS: The Peri 3 mm model achieved the best predictive performance among all radiomics models, with AUCs of 0.912 (95% CI: 0.836-0.961) and 0.856 (95% CI: 0.769-0.920) in the training and validation sets, respectively. The combined model integrating Peri 3 mm radiomics features, Log-AFP, and arterial phase peripheral enhancement showed AUCs of 0.948 (95% CI: 0.847-0.967) and 0.889 (95% CI: 0.728-0.934) in the training and validation sets. Exploratory evaluation in an independent testing set yielded consistent trends.

CONCLUSION: The combined model incorporating Peri 3 mm radiomics feature and clinical-radiological risk factors has significant predictive value for short-term prognosis after RFA, particularly for local tumor progression, offering potential to optimize follow-up protocols and enable timely clinical decision-making for high-risk patients. However, given the limited number of short-term poor prognosis events in the independent testing set, this set is an exploratory evaluation and requires external validation in multi-center, large-sample cohorts.