AUTHOR=Lu Chao , Pan Donggang , Shan Xiuhong , Dai Rao , Liu Bowen , Wang Zhixuan , Wang Xiaoxiao 

TITLE=Preoperative prognostic assessment using intratumoral and peritumoral adipose tissue radiomics derived from contrast-enhanced CT in cT3–4 gastric cancer

JOURNAL=Frontiers in Oncology

VOLUME=Volume 15 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2025.1631979

DOI=10.3389/fonc.2025.1631979

ISSN=2234-943X

ABSTRACT=PurposeExploring the value of contrast-enhanced computed tomography (CECT) based radiomics features from intratumoral and peritumoral adipose tissue (PAT) in predicting early recurrence (ER)after gastrectomy in patients with cT3–4 gastric cancer (GC).Materials and methodsThis retrospective study involved patients with cT3–4 GC who underwent preoperative CECT. The radiomics features of tumor and PAT were separately extracted from the CT venous phase images using the Pyradiomics package. The radiomic score (radscore) was computed for each patient by integrating LASSO regression-selected radiomic features, weighted according to their respective coefficients. The GC location, longest diameter, maximum thickness, cT stage and cN stage determined by preoperative CT were also evaluated. Univariate and bivariate analyses using the Cox regression model were performed to evaluate factors affecting ER. The Kaplan-Meier method was used for the analysis of ER-free survival.ResultsA total of 184 consecutive cT3–4 GC patients were enrolled in this study. Bivariate Cox regression analyses demonstrated that radscore and cT stage emerged as independent predictors of ER in all parameters. Radscore-based stratification showed a marked difference in the ER rates between high-risk patients (radscore ≥ -0.66) and low-risk patients (65.9% vs. 3.2%; log-rank p<0.001). Similarly, cT4 stage patients had markedly higher ER rates than cT3 stage patients (53.5% vs. 22.1%; log-rank p<0.001).ConclusionThe integrated radscore combining intratumoral and PAT features emerged as an independent prognostic predictor for ER in cT3–4 GC, offering quantitative biomarkers to optimize neoadjuvant therapy selection and postoperative surveillance intensity.