CT-based deep learning radiomics model for predicting proliferative hepatocellular carcinoma: application in transarterial chemoembolization and radiofrequency ablation

Abstract

Objectives: Proliferative hepatocellular carcinoma (HCC) is an aggressive tumor with varying prognosis depending on the different disease stages and subsequent treatment. This study aims to develop and validate a deep learning radiomics (DLR) model based on contrast-enhanced CT to predict proliferative HCC and to implement risk prediction in patients treated with transarterial chemoembolization (TACE) and radiofrequency ablation (RFA).

Materials and methods: 312 patients (mean age, 58 years ± 10 [SD]; 261 men and 51 women) with HCC undergoing surgery at two medical centers were included, who were divided into a training set (n = 182), an internal test set (n = 46) and an external test set (n = 84). DLR features were extracted from preoperative contrast-enhanced CT images. Multiple machine learning algorithms were used to develop and validate proliferative HCC prediction models in training and test sets. Subsequently, patients from two independent new sets (RFA and TACE sets) were divided into high- and low-risk groups using the DLR score generated by the optimal model. The risk prediction value of DLR scores in recurrence-free survival (RFS) and time to progression (TTP) was examined separately in RFA and TACE sets.

Results: The DLR proliferative HCC prediction model demonstrated excellent predictive performance with an AUC of 0.906 (95% CI 0.861–0.952) in the training set, 0.901 (95% CI 0.779–1.000) in the internal test set and 0.837 (95% CI 0.746–0.928) in the external test set. The DLR score effectively enables risk prediction for patients in RFA and TACE sets. For the RFA set, the low-risk group had significantly longer RFS compared to the high-risk group (P = 0.037). Similarly, the low-risk group showed a longer TTP than the high-risk group for the TACE set (P = 0.034).

Conclusions: The DLR-based contrast-enhanced CT model enables non-invasive prediction of proliferative HCC. Furthermore, the DLR risk prediction helps identify high-risk patients undergoing RFA or TACE, providing prognostic insights for personalized management.

Supplementary Information: The online version contains supplementary material available at 10.1186/s12880-025-01913-9.

Keywords: Contrast-enhanced CT; Deep learning radiomics; Hepatocellular carcinoma; Radiofrequency ablation (RFA); Transarterial chemoembolization (TACE).

Fig. 1

(A) Flowchart for patient inclusion and exclusion. (B) Workflow diagram for DLR model building and application. DLR = deep learning radiomics, HCC = hepatocellular carcinoma, RFA = radiofrequency ablation, TACE = transcatheter arterial chemoembolization

Fig. 2

DLR features screened by LASSO regression technique. (A) The figure shows the final selection of 23 non-zero coefficient features with high contribution features and the weight of each feature. (B) Distribution of regression coefficients for features

Fig. 3

Receiver operating characteristic curves of eight machine learning models based on DLR features in the training set and two test sets. The legend in the lower right corner of the figure shows the eight machine learning names. In the training set, internal and external test sets, the LR model was the one with the best performance. (A) Training set. (B) Internal test set. (C) External test set. AUC = area under the curve, LR = logistic regression, MLP = multi-layer perceptron, SVM = support vector machine

Fig. 4

Final DLR model DeLong test in different sets and Decision curves. The Delong test and Decision Curve Analysis (DCA) demonstrated superior performance and higher net clinical benefit of the DLR model. (A-B) Training set, (C-D) Internal test set, (E-F) External test set

Fig. 5

Survival analysis. In both RFA and TACE sets, patients in the low-risk group had better clinical outcomes. (A) Kaplan-Meier curve shows RFS difference in RFA set between high- and low-risk groups. (B) Kaplan-Meier curve shows TTP difference in TACE set between high- and low-risk groups. RFA = radiofrequency ablation, TACE = transcatheter arterial chemoembolization, TTP = time to progression, RFS = recurrence-free survival