AUTHOR=Wang Fen , Li Caiyun , Li Shuke , Zhang Teng , Yu Tongfu , Zhang Wei , He Jing , Yuan Mei , Gao Wen 

TITLE=CT-based radiomics integrated model for brain metastases in stage III/IV ALK-positive lung adenocarcinoma patients

JOURNAL=Frontiers in Oncology

VOLUME=Volume 15 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2025.1585930

DOI=10.3389/fonc.2025.1585930

ISSN=2234-943X

ABSTRACT=PurposeThis study aims to develop and validate a computed tomography (CT)-based radiomics nomogram for predicting brain metastases in lung adenocarcinoma with anaplastic lymphoma kinase positive (ALK+).MethodsOf 117 patients were retrospectively reviewed, among them, 34 patients from another hospital. Patients were randomly allocated into training (70%) and validation (30%) cohorts. We integrated the radiomics score (Rad_score) with independent clinic-radiological variables to build the nomogram model. The DeLong test and Decision curve analysis (DCA) were utilized to evaluate performance of three models. Cox regression analysis was used to identify statistically significant factors for progression-free survival (PFS) in ALK-positive lung adenocarcinoma, with model discrimination evaluated by the concordance index (C-index). The patients were divided into low-risk and high-risk groups. Finally, the Log-rank test was used to ascertain significant differences between the two risk groups in the nomogram models.ResultsFrom Stage III/IV lung cancer cases, we extracted 1834 radiomics features, identifying two features can serve as standalone indicators of BM. The AUC of radiomics model was 0.905 and 0.880 in the validation and external test cohort, respectively. The AUC of nomogram model was 0.940 in the validation cohort and 0.896 in the external test cohort, respectively. The statistical difference merely exists between nomogram and clinical model (P=0.009, P=0.012) in validation and external test cohorts, respectively. The multivariate Cox regression analysis showed that lymphadenopathy (Hazard ratio (HR) = 5.41, 95% confidence interval (CI): 1.38-21.16, P = 0.015) and rad_score (HR = 25.67, 95% CI: 5.41–121.94, P< 0.001) were independent predictive factors for PFS. The Concordance index (C-Index) for training cohort (C-Index(95%CI):0.887 (0.826-0.956); testing cohort:0.798 (0.676-0.938), and the external cohort with 0.927 (0.857-0.996). Patients in the low-risk group showed a significantly better PFS compared to those in the high-risk group in the training cohort and validation cohort (P all < 0.010, respectively), whereas the results were not consistent in the external test cohort (P=0.130).ConclusionCT-derived radiomic signatures show promise as a tool for predicting BM within 2 years after detection of primary lung adenocarcinoma detection with ALK+. Combing these radiomic signatures with clinical features can enhance risk stratification for these patients.