AUTHOR=Xia Xianwu , Liu Jinshuo , Du Yiman , Ma Xiaowen , Zhou Guoming , Yuan Jianjun , Hua Qianjin , Wang Lingling , Jiang Haitao , He Caidi , Liu Chibo 

TITLE=Fusion of CT radiomics and autoantibody biomarkers for enhanced prediction of lung cancer diagnosis: a comprehensive study

JOURNAL=Frontiers in Oncology

VOLUME=Volume 15 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2025.1591156

DOI=10.3389/fonc.2025.1591156

ISSN=2234-943X

ABSTRACT=IntroductionEarly and accurate diagnosis of lung cancer is crucial for improving treatment outcomes and patient survival. This study investigates the combined use of computed tomography (CT) radiomics and autoantibody biomarkers as a novel approach to enhance lung cancer diagnosis.MethodsWe analyzed 258 patients from two centers, dividing into training, internal validation, and external validation cohorts. CT scans were standardized, and 1106 radiomic features were extracted. The recursive feature elimination method was applied to iteratively eliminate the redundant features. Autoantibody levels were assessed using a multiplex immunoassay targeting seven specific biomarkers. After resampling the training dataset by using synthetic minority over-sampling technique, the support vector machine classifier was employed to train classification models. We developed separate predictive models for CT radiomics and autoantibody testing and then fused the two models to evaluate performance.ResultsThe fusion model demonstrated significantly improved diagnostic accuracy, with area under the receiver operating characteristic curve (AUC) values of 0.90 ± 0.02, 0.83 ± 0.08, and 0.78 ± 0.09 in three cohorts, outperforming both the CT radiomics-only (AUC: 0.87 ± 0.03, 0.76 ± 0.10, 0.74 ± 0.10) and autoantibody-only models (AUC: 0.67 ± 0.06, 0.55 ± 0.15, 0.57 ± 0.10). Decision curve analysis indicated a higher net benefit of the integrated model across various threshold probabilities.ConclusionThe fusion of CT radiomics and autoantibody biomarkers significantly enhances the diagnostic performance for lung cancer. This integrated approach enhances early detection and reduces unnecessary interventions, paving the way for personalized treatment strategies. Future research should focus on clinical validation and optimization of this model to facilitate its implementation in routine clinical practice.