Pretreatment multiparametric MRI-based habitat radiomics and 2.5D deep learning for predicting early complete response to induction chemoimmunotherapy in locally advanced nasopharyngeal carcinoma.
Authors
Affiliations (7)
Affiliations (7)
- Department of Radiation Oncology, Guangxi Medical University Cancer Hospital, Nanning, Guangxi 530021, People's Republic of China; Guangxi Clinical Medicine Research Center of Nasopharyngeal Carcinoma, Nanning, Guangxi 530021, People's Republic of China; Guangxi Key Laboratory of Early Prevention and Treatment for Regional High Frequency Tumor, Nanning, Guangxi 530021, People's Republic of China. Electronic address: [email protected].
- Department of Radiology, First Affiliated Hospital of Guangxi Medical University, Nanning, Guangxi 530021, People's Republic of China. Electronic address: [email protected].
- Department of Radiation Oncology, Guangxi Medical University Cancer Hospital, Nanning, Guangxi 530021, People's Republic of China; Guangxi Clinical Medicine Research Center of Nasopharyngeal Carcinoma, Nanning, Guangxi 530021, People's Republic of China; Guangxi Key Laboratory of Early Prevention and Treatment for Regional High Frequency Tumor, Nanning, Guangxi 530021, People's Republic of China. Electronic address: [email protected].
- Department of Radiation Oncology, Guangxi Medical University Cancer Hospital, Nanning, Guangxi 530021, People's Republic of China; Guangxi Clinical Medicine Research Center of Nasopharyngeal Carcinoma, Nanning, Guangxi 530021, People's Republic of China; Guangxi Key Laboratory of Early Prevention and Treatment for Regional High Frequency Tumor, Nanning, Guangxi 530021, People's Republic of China. Electronic address: [email protected].
- Department of Radiation Oncology, Guangxi Medical University Cancer Hospital, Nanning, Guangxi 530021, People's Republic of China; Guangxi Clinical Medicine Research Center of Nasopharyngeal Carcinoma, Nanning, Guangxi 530021, People's Republic of China; Guangxi Key Laboratory of Early Prevention and Treatment for Regional High Frequency Tumor, Nanning, Guangxi 530021, People's Republic of China. Electronic address: [email protected].
- Department of Radiology, First Affiliated Hospital of Guangxi Medical University, Nanning, Guangxi 530021, People's Republic of China. Electronic address: [email protected].
- Department of Radiation Oncology, Guangxi Medical University Cancer Hospital, Nanning, Guangxi 530021, People's Republic of China; Guangxi Clinical Medicine Research Center of Nasopharyngeal Carcinoma, Nanning, Guangxi 530021, People's Republic of China; Department of Oncology, Affiliated Wuming Hospital of Guangxi Medical University, Nanning, Guangxi, People's Republic of China; Guangxi Key Laboratory of Early Prevention and Treatment for Regional High Frequency Tumor, Nanning, Guangxi 530021, People's Republic of China. Electronic address: [email protected].
Abstract
To develop and externally validate a pretreatment multiparametric MRI-based combined imaging model integrating conventional radiomics, habitat radiomics, and 2.5D deep learning features for predicting early complete response (CR) after induction chemoimmunotherapy in locally advanced nasopharyngeal carcinoma (LANPC) METHODS: This retrospective study included 500 patients with LANPC who received induction chemoimmunotherapy between January 2021 and June 2025. Patients from one center were randomly divided into training and internal validation cohorts, and patients from another center formed an external test cohort. Pretreatment T1-weighted, T2-weighted fat-suppressed, and contrast-enhanced T1-weighted MRI were used to extract conventional radiomics, habitat radiomics, and 2.5D deep learning features. Clinical, conventional radiomics, deep learning, habitat radiomics, and combined imaging models were developed in the training cohort and evaluated using discrimination, calibration, decision curve analysis, and DeLong tests. Early CR was achieved in 51 of 195 patients in the training cohort, 25 of 84 in the internal validation cohort, and 59 of 221 in the external test cohort. The combined imaging model achieved the highest AUCs of 0.922, 0.847, and 0.821, respectively. In the external test cohort, the combined model achieved a sensitivity of 0.859 and a negative predictive value of 0.932. SHAP analysis indicated that both 2.5D deep learning and habitat radiomics features contributed prominently to model predictions. A pretreatment multiparametric MRI-based combined imaging model showed favorable performance for predicting early CR after induction chemoimmunotherapy in LANPC. This model may provide a noninvasive tool for pretreatment response stratification, although prospective multicenter validation is warranted.