Dual-energy computed tomography for predicting cervical lymph node metastasis in laryngeal squamous cell carcinoma.

RATIONALE AND OBJECTIVES

We constructed a dual-energy computed tomography (DECT)-based model to assess cervical lymph node metastasis (LNM) in patients with laryngeal squamous cell carcinoma (LSCC).

MATERIALS AND METHODS

We retrospectively analysed 164 patients with LSCC who underwent preoperative DECT from May 2019 to May 2023. The patients were randomly divided into training (n = 115) and validation (n = 49) cohorts. Quantitative DECT parameters of the primary tumours and their clinical characteristics were collected. A logistic regression model was used to determine independent predictors of LNM, and a nomogram was constructed along with a corresponding online model. Model performance was assessed using the area under the curve (AUC) and the calibration curve, and the clinical value was evaluated using decision curve analysis (DCA).

RESULTS

In total, 64/164 (39.0 %) patients with LSCC had cervical LNM. Independent predictors of LNM included normalized iodine concentration in the arterial phase (odds ratio [OR]: 8.332, 95 % confidence interval [CI]: 2.813-24.678,  < 0.001), normalized effective atomic number in the arterial phase (OR: 5.518, 95 % CI: 1.095-27.818,  = 0.002), clinical T3-4 stage (OR: 5.684, 95 % CI: 1.701-18.989,  = 0.005), and poor histological grade (OR: 5.011, 95 % CI: 1.003-25.026,  = 0.049). These predictors were incorporated into the DECT-based nomogram and the corresponding online model, showing good calibration and favourable performance (training AUC: 0.910, validation AUC: 0.918). The DCA indicated a significant clinical benefit of the nomogram for estimating LNM.

CONCLUSIONS

DECT parameters may be useful independent predictors of LNM in patients with LSCC, and a DECT-based nomogram may be helpful in clinical decision-making.