CT-based model predicts early progression after TACE in liver cancer

Patients with unresectable hepatocellular carcinoma who undergo transarterial chemoembolization (TACE) can face very different courses of disease, and clinicians currently have limited tools to identify, before treatment begins, who is most likely to progress early. A new retrospective multicenter study reports that a model combining pretreatment arterial-phase CT imaging features with clinical variables can predict time to progression after the first TACE session, offering a potential basis for individualized surveillance planning and timely reassessment of therapy.

The study, published in Cancer Imaging, drew on 508 patients across three cohorts. The public WAW-TACE dataset, comprising 226 patients, was used for model development and split into a training cohort of 158 patients and an internal validation cohort of 68 patients. Two institutional cohorts, cohort A with 252 patients and cohort B with 30 patients, served as independent external validation sets. Eligible patients had treatment-naive unresectable hepatocellular carcinoma treated with conventional TACE as initial therapy, preserved liver function, no extrahepatic metastasis, no major vascular invasion, and an available pretreatment contrast-enhanced CT scan.

Technically, the researchers built a time-to-progression prediction network, called TTP-Net, by integrating three kinds of information: clinical variables, radiomics features, and 2.5D deep learning features, all derived from arterial-phase CT images acquired before treatment.

In the reported results, TTP-Net achieved the highest observed performance among the compared models. It reached C-indices of 0.719 in the training cohort, 0.723 in the internal validation cohort, 0.713 in external cohort A, and 0.707 in external cohort B. The corresponding 12-month area under the curve values were 0.872, 0.862, 0.816, and 0.832, respectively. These figures indicate moderately strong discrimination, and the fact that performance held up in two external cohorts that the model never saw during development is a key strength of the evaluation.

Beyond the aggregate performance metrics, Kaplan-Meier analysis showed significant separation between patients classified as high risk and low risk in all four cohorts. This suggests the model could stratify individuals into groups with meaningfully different observed times to progression after TACE.

The authors conclude that integrating pretreatment arterial-phase CT-derived imaging features with clinical variables enabled identification of patients with hepatocellular carcinoma at high risk of early progression after the first TACE session. According to the study, such identification could support individualized surveillance planning and timely treatment reassessment after TACE, addressing a recognized gap, since pretreatment tools for flagging patients at high risk of early progression have remained limited despite the widespread use of TACE.

Several caveats are worth noting. The study is retrospective, and the model’s performance figures are observed in the specific cohorts analyzed; the external cohort B was small, with only 30 patients, so its estimates carry considerable uncertainty. The paper describes a planned analytical evaluation of a prediction model rather than a demonstration that using the model changes clinical decisions or improves patient outcomes. The article is published open access, and the version shared at publication is a peer-reviewed accepted manuscript that remains subject to further editorial edits before the final version of record.

Subject of Research: Prediction of time to progression after transarterial chemoembolization in hepatocellular carcinoma using a CT-based clinical and deep learning model

Article Title: Pretreatment arterial-phase CT-based prediction of time to progression after transarterial chemoembolization in hepatocellular carcinoma: a multicenter study

Article References: Wang, L., Xia, C., Peng, Z., Zhao, X., Wang, Y., Wu, Z., Lyu, Y., & Wu, R. (2026). Pretreatment arterial-phase CT-based prediction of time to progression after transarterial chemoembolization in hepatocellular carcinoma: a multicenter study. Cancer Imaging. https://doi.org/10.1186/s40644-026-01136-3

Image Credits: AI Generated

DOI: 10.1186/s40644-026-01136-3

Keywords: hepatocellular carcinoma, transarterial chemoembolization, arterial-phase CT, radiomics, deep learning, time to progression, prediction model, multicenter study, external validation, Cancer Imaging, Pretreatment, arterial-phase