DOI: 10.1097/js9.0000000000005583 ISSN: 1743-9159

Diagnostic efficacy of AI-driven radiomics in predicting lymph node metastasis in cholangiocarcinoma: a systematic meta-analysis

Haibo Xu, Yichen Wu, Haixing Fang, Zongting Gu
Show PDF Cite

Background:

Lymph node metastasis (LNM) critically impacts prognosis and surgical outcomes in cholangiocarcinoma (CCA), yet conventional imaging demonstrates limited preoperative detection accuracy. Artificial intelligence (AI)-driven radiomics offers a promising alternative; however, evidence in CCA remains nascent. This systematic meta-analysis evaluates the diagnostic performance of AI-driven radiomics for predicting LNM in CCA.

Methods:

Following PRISMA guidelines, we searched PubMed, Embase, Web of Science, and the Cochrane Library from inception through 31 December 2025. Eligible studies utilized CT- or MRI-based radiomics for LNM prediction in pathologically confirmed CCA, with histopathology as the reference standard. Pooled sensitivity, specificity, and AUC were calculated using Bayesian hierarchical models. Heterogeneity, publication bias, and subgroup analyses were performed, and methodological quality was assessed using the RQS and QUADAS-C tools.

Results:

Six retrospective studies (1023 patients, all from China) were included, of which four enabled quantitative synthesis. The pooled AUC was 0.88 (95% CI: 0.85–0.91), sensitivity was 0.86 (95% CI: 0.77–0.92), and specificity was 0.75 (95% CI: 0.67–0.82), with low heterogeneity ( I 2  < 25%) and no evidence of publication bias (Egger’s P = 0.63). AI-driven radiomics showed higher diagnostic performance than reported conventional imaging (false-negative rate of approximately 15%) and remained stable across cancer subtypes (intrahepatic vs. extrahepatic: AUC 0.89 vs. 0.88; P = 0.773). Fagan’s nomogram indicated an approximately 30% reduction in misdiagnosis.

Conclusions:

AI-driven radiomics shows promising diagnostic performance for LNM prediction in CCA and may have the potential to support surgical planning and postoperative management within NCCN/AJCC frameworks. However, the current evidence is constrained by retrospective designs, China-centric data reflecting regional epidemiology (higher CCA incidence related to cholelithiasis, liver fluke, and hepatitis B infections), a small number of included studies, and limited external validation. These findings should therefore be regarded as exploratory and hypothesis-generating, and require confirmation through prospective, multicenter studies in diverse populations before clinical adoption.

More from our Archive