医学
化学免疫疗法
无线电技术
列线图
鼻咽癌
完全响应
放射科
肿瘤科
内科学
放射治疗
癌
核医学
多参数磁共振成像
作者
Bifa Zhu,Kaihua Chen,Zuguang Wen,Liru Zhu,Ling Li,Xiaodong Zhu
标识
DOI:10.1016/j.ejrad.2026.113112
摘要
PURPOSE: To develop and externally validate a pretreatment multiparametric MRI-based habitat radiomics nomogram for predicting early complete response (CR) after induction chemoimmunotherapy in locally advanced nasopharyngeal carcinoma (LANPC). MATERIAL AND METHODS: This two-center retrospective study included 435 patients with LANPC treated with induction chemoimmunotherapy between January 2021 and January 2025. Patients from Guangxi Medical University Cancer Hospital (GXMUCH) were divided into training (n = 195) and internal validation (n = 49) cohorts, and patients from the First Affiliated Hospital of Guangxi Medical University (FAH-GXMU) formed an independent external test cohort (n = 191). Pretreatment T1-weighted, fat-suppressed T2-weighted, and contrast-enhanced T1-weighted images were analyzed. Voxel-wise multiparametric MRI features were clustered to generate habitat subregions. Conventional radiomics, habitat radiomics, combined radiomics, and radiomics nomogram models were developed and evaluated using AUC, calibration, decision curve analysis, and DeLong test. RESULTS: Early CR was achieved in 51, 11, and 56 patients in the three cohorts, respectively. Habitat radiomics outperformed conventional whole-tumor radiomics, with the external test AUC increasing from 0.735 to 0.821. The combined radiomics model and radiomics nomogram achieved external test AUCs of 0.827 and 0.840, respectively. The nomogram yielded the highest numerical AUCs in the three cohorts (0.961, 0.888, and 0.840, respectively). CONCLUSION: Pretreatment multiparametric MRI-based habitat radiomics provided additional predictive information beyond conventional whole-tumor radiomics for early CR prediction after induction chemoimmunotherapy in LANPC. The radiomics nomogram may serve as a clinically interpretable tool for individualized response stratification, although prospective multicenter validation is warranted.
科研通智能强力驱动
Strongly Powered by AbleSci AI