Fractal analysis based on Gd-EOB-DTPA-enhanced MRI for prediction of vessels that encapsulate tumor clusters in patients with hepatocellular carcinoma

Objective: The aim of this study was to assess the potential role of fractal analysis derived from Gd-EOB-DTPA-enhanced MRI in predicting vessels that encapsulate tumor clusters (VETC) in patients with hepatocellular carcinoma (HCC). Methods: This retrospective study included 505 patients with HCC who underwent Gd-EOB-DTPA-enhanced MRI before surgical resection at two medical centers (training set: 253 patients, internal test set: 108 patients, external test set: 144 patients). The fractal dimension (FD) and lacunarity were extracted from the hepatobiliary phase of the tumor using box-counting algorithms. Additionally, conventional imaging features were evaluated. Univariate and multivariate logistic regression analyses were conducted in the training set to identify independent predictors for VETC, and a nomogram was created to visualize the final predictive model. The performance of these models was tested in the internal and external test sets. Recurrence-free survival (RFS) and overall survival (OS) were analyzed using the Kaplan-Meier method along with the log-rank test. Results: VETC-positive HCC exhibited higher FD and lacunarity than VETC-negative HCC (P < 0 .001). The FD-lacunarity model achieved an area under receiver operating characteristics curve (AUC) of 0.78 (95% confidence interval [CI]: 0.70-0.87) in the internal test set and 0.79 (95%CI: 0.70-0.86) in the external test set. Multivariate logistic regression analysis identified serum alpha-fetoprotein, tumor size, intratumor artery, FD, and lacunarity as independent predictors for VETC, which were used for constructing the hybrid model. A clinical model was established using AFP, tumor size, and intratumor artery alone. The diagnostic performance of the hybrid model was significantly surpassed that of the clinical-radiological model when fractal parameters were incorporated, with AUCs increasing from 0.72 to 0.80 in the internal test set and from 0.65 to 0.84 in the external test set (all P < 0.05). Patients predicted by the hybrid model to have VETC-positive HCC exhibited significantly shorter RFS and OS compared to those predicted to have VETC-negative HCC (P < 0.05). Conclusion: Fractal analysis based on Gd-EOB-DTPA-enhanced MRI enabled the quantitative characterization of VETC status by fractal dimension and lacunarity. The hybrid model may assist in estimating VETC and stratifying prognosis in patients with HCC.