Radiologist-guided multimodal fusion network for differentiating parotid tumors on 3D non-contrast CT
4 affiliations
- 1Department of Radiation Oncology, Precision Radiation in Oncology Key Laboratory of Sichuan Province, Sichuan Clinical Research Center for Cancer, Sichuan Cancer Hospital & Institute, Sichuan Cancer Center, University of Electronic Science and Technology of China
- 2Department of Radiation Oncology, Radiation Oncology Key Laboratory of Sichuan Province, Sichuan Clinical Research Center for Cancer, Sichuan Cancer Hospital & Institute, Sichuan Cancer Center, University of Electronic Science and Technology of China
- 3Department of Oncology, Meizhou People’s Hospital
- 4Department of Oncology, Bazhong City Hospital of Traditional Chinese Medicine
Abstract
Objectives: To improve the diagnostic accuracy of benign versus malignant parotid tumors using non-contrast CT. We propose a clinician-in-the-loop hybrid framework that mitigates low soft-tissue contrast by fusing semi-automated 3D visual features with expert semantic descriptors.
Methods: This multi-center retrospective study analyzed non-contrast CTs from 367 patients. We developed AttenNet, a semi-automated model. First, an nnU-Net module automates 3D tumor segmentation, eliminating manual bounding boxes. A 3D-ResNet backbone then extracts volumetric features, which are dynamically fused with radiologist-provided semantic labels via a Multi-Head Cross-Attention Fusion (MHAF) mechanism. This allows the network to query visual morphology based on expert priors. Diagnostic performance and Gradient-weighted Class Activation Mapping (Grad-CAM) interpretability were evaluated.
Results: The segmentation achieved a median Dice of 0.827. AttenNet (3D-ResNet) yielded an AUC of 0.908, significantly outperforming the image-only 3D-ResNet (AUC: 0.741) and expert-feature models (AUC: 0.821). External validation ( n = 20) reached an AUC of 0.560–0.910.However, given the extremely small sample size, these results must be interpreted strictly as an exploratory proof-of-concept rather than definitive confirmation of generalizability. Grad-CAM confirmed that AttenNet highlights clinically essential geometric indicators-specifically tumor margins and internal heterogeneity-for differentiating malignancy, whereas single-modality baselines focused on background artifacts.
Conclusion: This clinician-in-the-loop paradigm successfully integrates spatial imaging with expert reasoning. While multimodal fusion drove the primary performance gains, the MHAF mechanism provided enhanced spatial interpretability.
Related
- The radiologist’s dilemma: Differentiating non-ischemic pneumatosis intestinalis from ischemic bowel necrosis
- Quantitative Pharmacokinetic Analysis of Dynamic Contrast-Enhanced Perfusion in Multiparametric Prostate MRI: An Exploratory Intra-Patient Study Differentiating Malignant Lesions from Normal Tissue
- Differentiating pancreatic acinar cell carcinoma from solid pseudopapillary neoplasm using contrast-enhanced computed tomography
- Qualitative Dynamic Contrast-Enhanced MRI Time–Intensity Curve Analysis for Differentiating Non-Progressive Treatment-Related Changes from Tumor Progression After Stereotactic Radiotherapy for Brain Metastases
- Visualising uterine microcirculation using contrast-enhanced ultrasound for differentiating between healthy controls and participants with adenomyosis: rationale and design of the UteroVue Adenomyosis study.
BibTeX
@article{yang2026radiologistguided,
title = {{Radiologist-guided multimodal fusion network for differentiating parotid tumors on 3D non-contrast CT}},
author = {Qin Yang and Wenjun Liao and Jianda Sun and Dacheng Mu and Minyang Yang and Qiaohua Li and Yue Zhao and Shichuan Zhang},
journal = {Frontiers in Oncology},
year = {2026},
month = oct,
doi = {10.3389/fonc.2026.1926379},
url = {https://www.frontiersin.org/articles/10.3389/fonc.2026.1926379/full},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Radiologist-guided multimodal fusion network for differentiating parotid tumors on 3D non-contrast CT
AU - Qin Yang
AU - Wenjun Liao
AU - Jianda Sun
AU - Dacheng Mu
AU - Minyang Yang
AU - Qiaohua Li
AU - Yue Zhao
AU - Shichuan Zhang
PY - 2026
DA - 2026/10/15
JO - Frontiers in Oncology
DO - 10.3389/fonc.2026.1926379
UR - https://www.frontiersin.org/articles/10.3389/fonc.2026.1926379/full
AB - Objectives: To improve the diagnostic accuracy of benign versus malignant parotid tumors using non-contrast CT. We propose a clinician-in-the-loop hybrid framework that mitigates low soft-tissue contrast by fusing semi-automated 3D visual features with expert semantic descriptors. Methods: This multi-center retrospective study analyzed non-contrast CTs from 367 patients. We developed AttenNet, a semi-automated model. First, an nnU-Net module automates 3D tumor segmentation, eliminating manual bounding boxes. A 3D-ResNet backbone then extracts volumetric features, which are dynamically fused with radiologist-provided semantic labels via a Multi-Head Cross-Attention Fusion (MHAF) mechanism. This allows the network to query visual morphology based on expert priors. Diagnostic performance and Gradient-weighted Class Activation Mapping (Grad-CAM) interpretability were evaluated. Results: The segmentation achieved a median Dice of 0.827. AttenNet (3D-ResNet) yielded an AUC of 0.908, significantly outperforming the image-only 3D-ResNet (AUC: 0.741) and expert-feature models (AUC: 0.821). External validation ( n = 20) reached an AUC of 0.560–0.910.However, given the extremely small sample size, these results must be interpreted strictly as an exploratory proof-of-concept rather than definitive confirmation of generalizability. Grad-CAM confirmed that AttenNet highlights clinically essential geometric indicators-specifically tumor margins and internal heterogeneity-for differentiating malignancy, whereas single-modality baselines focused on background artifacts. Conclusion: This clinician-in-the-loop paradigm successfully integrates spatial imaging with expert reasoning. While multimodal fusion drove the primary performance gains, the MHAF mechanism provided enhanced spatial interpretability.
SN - 2234-943X
ER -