A Dual-Center, Interpretable Deep Learning-Radiomics Fusion Model for Distinguishing Nasopharyngeal…

AI / informaticsAcademic radiology· Aug 2026

A Dual-Center, Interpretable Deep Learning-Radiomics Fusion Model for Distinguishing Nasopharyngeal Carcinoma from Nasopharyngeal Lymphoma Using FDG PET/CT.

MeiNa Liang, ChengMao Guo, Xiaoyu Gao … JingXing Xiao

Medical Imaging AIAI / informaticsDiagnosis

A deep learning-radiomics model accurately distinguishes nasopharyngeal carcinoma from lymphoma.

A Dual-Center, Interpretable Deep Learning-Radiomics Fusion Model for Distinguishing Nasopharyngeal Carcinoma from Nasopharyngeal Lymphoma Using FDG PET/CT.

MeiNa Liang … JingXing Xiao

Academic radiology · Aug 2026

Background

Accurate differentiation between Nasopharyngeal Carcinoma (Nasopharyngeal Carcinoma) and Nasopharyngeal Lymphoma (Nasopharyngeal Lymphoma) is critical for clinical management.

Purpose

We aimed to develop an interpretable deep learning-radiomics fusion model using positron emission tomography/computed tomography (PET/CT) and validate it externally.

Methods

In this retrospective dual-center study, we enrolled 434 patients (349 internal, 85 external) with pathologically confirmed, treatment-naive Nasopharyngeal Carcinoma or Nasopharyngeal Lymphoma who underwent baseline FDG PET/CT.

Results

AUC of 0.833

model accurately distinguishes nasopharyngeal carcinoma from lymphoma (AUC; 1.0 is perfect)

More results

Among them, 329 patients (75.8%) had Nasopharyngeal Carcinoma ( internal: 266, external: 63) and 105 patients (24.2%) had Nasopharyngeal Lymphoma( internal: 83, external: 22).

The mean age of Nasopharyngeal Carcinoma patients was 50.5 ± 13.8 years, while that of Nasopharyngeal Lymphoma patients was 59.0 ± 16.5 years.

More results

SHAP analysis identified key predictors: original_shape_MinorAxisLength_PET and lbp-3D-k_glrlm_RunLengthNonUniformity_PET for Nasopharyngeal Lymphoma, and wavelet-LLL_firstorder_Maximum_CT features for Nasopharyngeal Carcinoma.

“

Conclusion

We developed a robust, interpretable fusion model for Nasopharyngeal Carcinoma/Nasopharyngeal Lymphoma differentiation, offering high specificity and transparent decision logic, with potential as a non-invasive clinical decision-support tool.

A Dual-Center, Interpretable Deep Learning-Radiomics Fusion Model for Distinguishing Nasopharyngeal Carcinoma from Nasopharyngeal Lymphoma Using FDG PET/CT.

A Dual-Center, Interpretable Deep Learning-Radiomics Fusion Model for Distinguishing Nasopharyngeal Carcinoma from Nasopharyngeal Lymphoma Using FDG PET/CT.

Related papers