AI and Cancer Research ── Liver ── 2026-11-02
Liver cancer: treatment, trials and AI
- Abstract
Hepatocellular carcinoma (HCC) remains a leading cause of cancer-related mortality worldwide, with high recurrence rates following curative resection. Accurate non-invasive assessment of tumor aggressiveness and recurrence risk is critical for surgical decision-making and postoperative surveillance, yet remains challenging with conventional imaging modalities. Here, we developed a CT lesion foundation model (LFM), trained via self-supervised learning on 32,120 generic lesion and 6,387 HCC-specific CT patches to capture transferable intra-lesional imaging phenotypes. Across one internal and two external validation cohorts, the sequentially pretrained LFM (LFM-Seq) achieved area under the curve values of 0.916, 0.846, and 0.920 for microvascular invasion (MVI) assessment, and 0.833, 0.885 and 0.795 for pathological grade assessment. The model remained robust under scenarios of limited training data and tumor localization perturbations. Furthermore, LFM-Seq-derived risk scores stratified recurrence-free survival (RFS), yielding an externally validated C-index of 0.701, and remained independently associated with RFS after adjusting for clinicopathological factors. Exploratory biological interpretability analyses showed that LFM-Seq-derived risk phenotypes were associated with suppressed lymphocyte-related immune programs. These results suggest that CT lesion foundation models can support non-invasive and robust characterization of HCC aggressiveness, recurrence risk stratification and discovery of interpretable prognostic imaging biomarkers.
Journal IF-equivalent: 9.5 (OpenAlex 2-year mean citedness, value as of 2026-10-18, retrieved 2026-10-19; not the official Clarivate IF)Reference: Xia W, Shen Z, Li M, Lu J, Xia J, Cheng Y, et al. A CT Lesion Foundation Model for Non‐Invasive Assessment of Hepatocellular Carcinoma Aggressiveness and Prediction of Recurrence Risk. Advanced Science. 2026 Nov 1 [Epub ahead of print]. doi:10.1002/advs.78587.Checked: Abstract only - Abstract
Introduction: We aimed to evaluate the possible association between TNFRSF4 and the prognosis of hepatocellular carcinoma (HCC), and particularly to develop CT-based radiomics models to predict TNFRSF4 levels before surgery with either tumor per se or peritumoral changes.
Methods: 286 HCC patients' genomic data from the Cancer Genome Atlas (TCGA) and 35 individuals' arterial phase enhanced computed tomography (CT) data from the Cancer Imaging Archive (TCIA) were used for the prognostic analysis and radiomics feature extraction. Accordingly, the machine-learning algorithms, including maximum relevance minimum redundancy (mRMR) and recursive feature elimination (RFE), were employed to construct the whole-tumor and whole-tumor-peritumoral (whole tumor and a 3 mm peritumor area) models with the receiver operating characteristic curve evaluating the accuracy. Results & Conclusions High TNFRSF4 expression was associated with shorter overall survival in this cohort. For the whole tumor model, 20 optimal features were extracted with mRMR, followed by screening out 3 representative subsets with RFE. When combined with the logistic regression model, the area under the receiver operating characteristic curve (AUC) for the training and verification sets equaled 0.813 and 0.759, respectively. Subsequently, the whole-tumor-peritumoral model was established, as well as 20 features derived from the mRMR algorithm and 4 typical subsets from RFE, and it was found that the AUC value was 0.871 in the training set and 0.803 in the verification set. In conclusion, the radiomics model based on the tumor per se and peritumoral information predicted the TNFRSF4 expression and its association with overall survival in HCC patients.
Journal IF-equivalent: 9.5 (OpenAlex 2-year mean citedness, value as of 2026-10-18, retrieved 2026-10-19; not the official Clarivate IF)Reference: Ye R, Wang S, Li Y, He Q. The prognostic value of TNFRSF4 in hepatocellular carcinoma using a two-zone CT imaging model. Heliyon. 2026 Nov;12(16):e45637. doi:10.1016/j.heliyon.2026.e45637.Checked: Abstract only