One-stop dual-energy perfusion CT in predicting pathological differentiation of laryngeal and hypopharyngeal squamous cell carcinoma

To evaluate the predictive efficacy of parameters derived from one-stop dual-energy perfusion CT and multiparameter combined models for assessing the differentiation degree of laryngeal and hypopharyngeal squamous cell carcinoma (LHSCC). 82 pathologically confirmed LHSCC patients who underwent preoperative one-stop dual-energy perfusion CT were retrospectively enrolled. The measured parameters included the CT values of the lesion in 74 keV monochromatic images, iodine concentration (IC), standardized iodine concentration (sIC), effective atomic number (Zeff), standardized effective atomic number (sZeff), spectral curve slope (λ HU ), blood volume (BV), blood flow (BF), mean transit time (MTT), permeability surface area product (PS), and time-to-maximum (Tmax). Differences in these parameters among the groups with well, moderate, and poor differentiation were compared. Independent predictors were screened via logistic regression to construct combined spectral-perfusion models. Diagnostic performance was evaluated using receiver operating characteristic (ROC) curves, and agreement between the combined models and pathological diagnosis was assessed using Cohen's kappa tests. Among the 82 cases of LHSCC, 16 (19.51%) were well-differentiated, 48 (58.54%) were moderately differentiated, and 18 (21.95%) were poorly differentiated. Significant differences were observed in arterial and venous phase spectral parameters (IC, sIC, Zeff, sZeff, λ HU ) and perfusion parameters (BF, BV, PS, MTT) across the three groups (all p < 0.05). The combined model Y2, constructed using the independent predictors of arterial phase IC, venous phase IC, venous phase sZeff, BV, and MTT, achieved an AUC of 0.938 for differentiating well- from poorly differentiated LHSCC. The combined model Y3, based on the independent predictors of arterial phase IC and BV, yielded an AUC of 0.915 for distinguishing moderately from poorly differentiated LHSCC. The spectral-perfusion combined models demonstrated substantial agreement with pathological diagnoses (well/poor differentiation: κ = 0.766; moderate/poor differentiation: κ = 0.705; both p < 0.05), with overall concordance rates of 88.235% and 87.879%, respectively. Spectral parameters (arterial/venous phase IC, sIC, Zeff, sZeff, λ HU ) and perfusion parameters (BF, BV, PS, MTT) obtained from one-stop dual-energy perfusion CT can effectively evaluate LHSCC differentiation. Moreover, the spectral-perfusion multiparameter combined models exhibit higher predictive value.