Related Experiment Video
Updated: Sep 13, 2026
Utilizing 18F-FDG PET/CT Imaging and Quantitative Histology to Measure Dynamic Changes in the Glucose Metabolism in Mouse Models of Lung Cancer
Published on: July 21, 2018
Non-invasive radiogenomic assessment of growth-associated protein 43 in lung adenocarcinoma brain metastases
Chen Sun1, Yuqi Liu1, Chenggang Jiang1
1The Department of Neurosurgery, Beijing Tiantan Hospital, Capital Medical University, No. 119 West Road, South Fourth Ring Road, Beijing 100070, China.
Background:
To evaluate the prognostic role of growth-associated protein 43 (GAP43) in lung adenocarcinoma (LUAD) brain metastases and develop a non-invasive radiogenomic model combining MRI radiomics with transcriptomics for GAP43 prediction and survival estimation.
Methods:
This retrospective study included 308 patients with surgically resected LUAD brain metastases (2012-2021). Fifty-six patients with paired RNA sequencing and contrast-enhanced T1-weighted MRI (CE-T1WI) were used to build the radiomics model predicting GAP43 expression. An additional independent cohort of 252 patients was included for exploratory prognostic evaluation of the imaging-derived radiomics score (RS). Radiomic features (n = 1,130) were extracted using PyRadiomics and underwent sequential feature selection, including inter-observer reproducibility assessment using intraclass correlation coefficients (ICCs), univariate t-tests, correlation-based redundancy filtering, and least absolute shrinkage and selection operator (LASSO)-based feature selection. An optimized support vector machine (SVM) model was subsequently developed to predict GAP43 expression and generate the RS. Survival was analyzed via Kaplan-Meier and Cox regression.
Results:
A total of 56 patients were enrolled in the radiogenomics cohort. GAP43 was significantly independently associated with poorer overall survival (OS; P = 0.026). High GAP43 expression correlated positively with activated mast cells (R = 0.53), eosinophils (R = 0.34). The radiogenomic model showed favorable performance for predicting GAP43 expression in repeated nested cross-validation, with an AUC of 0.825 (95% CI, 0.791-0.858). In an independent prognostic cohort of 252 patients, a higher imaging-derived radiomics score (RS) was independently associated with poorer OS (P < 0.001). The RS-based nomogram showed acceptable prognostic performance, with time-dependent AUCs of 0.661, 0.666, and 0.631 for 3-, 6-, and 12-month OS, respectively.
Conclusions:
GAP43 is an adverse prognostic biomarker in LUAD brain metastases. A CE-T1WI radiogenomic model accurately predicts GAP43 expression, facilitating individualized short-term survival stratification.
