Deep Learning Model Based on Dual-energy CT for Assessing Cervical Lymph Node Metastasis in Oral Squamous Cell
Yi-Ming Qi1, Li-Jie Zhang1, Yu Wang2
1Department of Radiology, The Second Xiangya Hospital of Central South University, Changsha, China (Y.M.Q., L.J.Z., Y.J.L., E.H.X., Y.H.L.).
Academic Radiology
|July 8, 2025
Summary
A new deep learning model accurately detects lymph node metastasis (LNM) in oral squamous cell carcinoma (OSCC) using dual-energy CT scans. This AI approach shows promise for improving patient treatment planning and clinical decisions.
Area of Science:
- Radiology
- Oncology
- Artificial Intelligence
Background:
- Accurate detection of lymph node metastasis (LNM) is critical for oral squamous cell carcinoma (OSCC) treatment planning.
- Dual-energy CT (DECT) offers advanced imaging capabilities for LNM assessment.
Purpose of the Study:
- To develop and validate a deep learning model for enhanced LNM detection in OSCC using DECT.
- To compare the model's performance against human radiologists.
Main Methods:
- DECT images (Iodine Map, Fat Map, 70 keV monoenergetic, RHO/Z Map) and clinical data from two centers were used.
- Fused four-channel composite images were created by stacking region-of-interest images.
- 16 deep learning models were trained, including Crossformer, Densenet169, Squeezenet1_0, and a novel Crossformer_Transformer model.
Main Results:
- The Crossformer_Transformer model achieved the highest AUC of 0.960 (training), 0.881 (internal validation), and 0.881 (external validation).
- The model outperformed radiologists, whose average AUC ranged from 0.723 to 0.819.
- The model demonstrated robust diagnostic performance on multicenter data.
Conclusions:
- The Crossformer_Transformer model shows significant potential for improving preoperative risk assessment of cervical LNM in OSCC.
- This AI-driven approach can enhance clinical decision-making for OSCC patients.
