Value of intratumoral combined with peritumoral radiomics model in the prediction of Ki-67 expression in lung adenocarcinoma: a multi-center study

Background Ki-67 proliferation index shows strong association with tumor aggressiveness, while non-invasive quantification approaches remain limited in clinical applications. The study assessed the role played by intratumoral and peritumoral radiomics together with radiomics nomogram in Ki-67 expression prediction in lung adenocarcinoma (LUAD). Methods A cohort of 388 individuals diagnosed with LUAD at the Fifth Affiliated Hospital of Wenzhou Medical University (WMU) fell into a training cohort (TC, n = 272) and an internal validation cohort (IVC, n = 116). Furthermore, 176 and 221 patients from the First Affiliated Hospital of Jiaxing University and the Sixth Affiliated Hospital of WMU constituted external validation cohort 1 (EVC1) and 2 (EVC2), respectively. We extracted 1,688 CT-based radiomics features from the gross tumor volume (GTV) encompassing peritumoral areas of 3, 6, and 9 mm (GPTV3, GPTV6, GPTV9). The study constructed a radiomics nomogram, which combined independent predictors in clinical practice with the radiomics score (Rad-score) generated by the most effective radiomics model. Results The GPTV6 radiomics model with AUC values of 0.896, 0.844, 0.784, and 0.835 in the TC, IVC, EVC1 and EVC2, respectively. The delong test shows that it outperformed the GTV, GPTV3, and GPTV9 models in relevant prediction. In the clinical model, independent predictors included nodule size, spicule sign, and nodule density. The combined nomogram, incorporating these predictors and the GPTV6-Radscore, demonstrated significant clinical utility, achieving AUCs of 0.905, 0.871, 0.806, and 0.882 across the respective cohorts. Conclusions CT-based radiomics nomogram excels in Ki-67 expression prediction in patients suffering from LUAD, providing reference for the personalized treatment of these patients. Trial registration: This study was retrospectively registered by the Ethics Committee of Wenzhou Medical University Affiliated Fifth Hospital (2023 − 636), and the requirement for informed consent was waived for all participants.