Diagnostic performance of machine learning-based radiomics models for predicting epidermal growth factor receptor mutation status in lung adenocarcinoma in Chinese patients: A systematic review and meta-analysis

Clinical Snapshot

Key Findings

• P Value: Deep learning vs. conventional ML: P=0.012; CT vs. PET-CT: P=0.038; external vs. internal validation: P=0.006; imaging modality as heterogeneity source: P<0.05

• Effect Size: Summary AUC 0.85; pooled sensitivity 71%; pooled specificity 81%. Deep learning AUC 0.871 vs. conventional ML AUC 0.798. CT-based AUC 0.879 vs. PET-CT AUC 0.828. External validation AUC 0.922 vs. internal validation AUC 0.841.

• Primary Outcome: Diagnostic accuracy of ML-based radiomics models for predicting EGFR mutation status in Chinese lung adenocarcinoma patients

• Nnt Or Sensitivity: Pooled sensitivity 71% (95% CI 68–74%); pooled specificity 81% (95% CI 78–84%); positive likelihood ratio and negative likelihood ratio not reported in abstract. At 71% sensitivity, approximately 29% of EGFR-mutant cases would be missed if used as a standalone triage tool.

• Confidence Interval: sAUC 95% CI: 0.82–0.88; sensitivity 95% CI: 68–74%; specificity 95% CI: 78–84%

Clinical Application

CT-based radiomics is technically feasible in most tertiary oncology centres with access to dedicated radiomics software. However, standardisation of image acquisition protocols, segmentation workflows, and model deployment pipelines remains a significant barrier to routine clinical implementation. Deep learning models with external validation represent the most clinically credible configuration based on this review's findings. In Australia, EGFR mutation testing is standard of care for all patients with advanced non-squamous NSCLC prior to first-line therapy, supported by RACGP and Cancer Australia guidelines. EGFR-targeted therapies (osimertinib, gefitinib, erlotinib, afatinib) are PBS-listed for EGFR-mutant NSCLC. The EGFR mutation prevalence in Australian patients of East Asian descent (~40–50%) is substantially higher than in non-Asian Australians (~10–15%), meaning the diagnostic performance metrics from this Chinese cohort are most directly applicable to Asian-Australian patients. TGA has not approved any radiomics-based diagnostic tool for EGFR prediction as of 2025. This technology should be considered investigational in the Australian context. Liquid biopsy (ctDNA) remains a clinically validated non-invasive alternative when tissue is insufficient, and is increasingly available through Australian genomics services. Chinese (East Asian) patients with confirmed lung adenocarcinoma on CT or PET-CT imaging who are candidates for EGFR mutation testing. Most directly applicable to patients where tissue biopsy is technically challenging, contraindicated, or where rapid triage prior to molecular testing is clinically desirable.

Abstract

ObjectiveThis systematic review and meta-analysis evaluates the diagnostic performance of machine learning-based radiomics models for predicting epidermal growth factor receptor mutation status in Chinese patients with lung adenocarcinoma.MethodsFollowing the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 guidelines and prospectively registered in the International Prospective Register of Systematic Reviews (CRD420251273027), a systematic search of PubMed, Embase, Web of Science, the Cochrane Library, Scopus, China National Knowledge Infrastructure, Wanfang, VIP, and Chinese Biomedical Literature Database was conducted from inception to 31 October 2025. Two reviewers independently screened studies, extracted data, and assessed bias using the Quality Assessment of Diagnostic Accuracy Studies-2 tool. A bivariate random-effects model was used to synthesize the data. Subgroup analyses were conducted for three factors: (a) imaging modality (computed tomography vs. positron emission tomography-computed tomography); (b) algorithm type (deep learning vs. conventional machine learning); and (3) validation strategy (external vs. internal).ResultsThirteen studies encompassing 6628 patients were included. The pooled sensitivity was 71% (95% confidence interval: 68-74), the pooled specificity was 81% (95% confidence interval: 78-84), and the summary area under the curve was 0.85 (95% confidence interval: 0.82-0.88). Deep learning models significantly outperformed conventional machine learning models (area under the curve: 0.871 vs. 0.798; P = 0.012). Computed tomography-based models yielded higher accuracy than positron emission tomography-computed tomography-based models (area under the curve: 0.879 vs. 0.828; P = 0.038). Models validated on independent external cohorts demonstrated superior performance compared with those relying solely on internal validation (area under the curve: 0.922 vs. 0.841; P = 0.006). Imaging modality was a significant source of heterogeneity (P < 0.05). No threshold effect or publication bias was detected.ConclusionMachine learning-based radiomics models exhibit promising diagnostic accuracy for the noninvasive prediction of epidermal growth factor receptor mutations in Chinese patients with lung adenocarcinoma. Computed tomography-based deep learning models subjected to independent external validation represent the current optimal approach. However, the retrospective nature and substantial heterogeneity of the included studies necessitate large-scale, prospective, multicenter trials with standardized workflows before clinical translation.

References

1. 1.Yang, J., Jiang, J., Peng, J., Li, J., Mi, P., & Chen, G. (2026). Diagnostic performance of machine learning-based radiomics models for predicting epidermal growth factor receptor mutation status in lung adenocarcinoma in Chinese patients: A systematic review and meta-analysis. The Journal of International Medical Research. https://doi.org/10.1177/03000605261460311