Radiomics features combined with clinical and CT features for predicting the Ki-67 index in stage T1 non-small cell lung cancer patients: a multicenter study.

Background: Non-small cell lung cancer (NSCLC) is a major global public health concern due to its high mortality and late-stage detection. The Ki-67 index, a key marker of tumor proliferation, is crucial for evaluating tumor aggressiveness and guiding personalized treatment. However, current detection relies on invasive surgical histology or biopsies, which are limited by tumor heterogeneity and potential misdiagnosis. Radiomics enables non-invasive extraction of quantitative features from computed tomography (CT) images, offering a promising solution for predicting Ki-67 expression. Existing radiomics studies for NSCLC mainly focus on intratumoral features, ignoring peritumoral regions and clinical/CT predictors, which limits clinical utility. This study aimed to develop and validate a combined model integrating intratumoral/peritumoral radiomics features with clinical and CT factors for predicting the Ki-67 index in stage T1 NSCLC patients. Methods: A retrospective multicenter study was conducted, with 247 eligible patients from three centers (January 2020 to August 2024) divided into a training set (70%, n=172) and an external validation set (n=75 from two additional centers). To assess the effectiveness of each model in predicting the Ki-67 index status, receiver operating characteristic (ROC) curves were produced and the area under the curve (AUC) was computed. The DeLong test was used to compare the areas under two related ROC curves. Decision curve analysis (DCA) was performed to assess the clinical utility of each model, while the calibration curve was utilized to assess the level of calibration for each model. The best radiomics features of the intratumour (INTRA), peritumour 3 mm (Peri3 mm), intratumour and peritumour 3 mm (IntraPeri3 mm) and image fusion of the intratumour and peritumoural 3 mm (ImageFusion3 mm) areas were extracted and screened to construct radiomics models (model INTRA, model Peri3 mm, model IntraPeri3 mm and model ImageFusion3 mm), and the optimal radiomics model was screened. A nomogram was generated after a combination model was built using the optimal model radiomics score and independent clinical and CT predictors. Results: Gender, density, and lobulation were found to be associated with elevated Ki-67 expression. The best radiomics model was thought to be the IntraPeri3 mm model. In comparison to the clinical and IntraPeri3 mm models, the combined model had a higher AUC. The combined model, the IntraPeri3 mm model, and the clinical model all had high calibration degrees. At the threshold of 0.10-0.80, the combined model had a significant positive impact on clinical outcomes. Conclusions: The Ki-67 index can be predicted by combining intratumoural and peritumoural radiomics data with clinical and CT features.