Non-contrast dual-layer spectral detector CT combined with laboratory indicators for differentiating hepatic hemangiomas from malignant liver tumours: development and internal validation of a diagnostic prediction model
Year: 2026
Type: article
Study design: Observational Study
Abstract: Background: Hepatic hemangiomas, hepatocellular carcinoma (HCC), and metastatic liver cancer (MLC) may all appear hypoattenuating or isoattenuating on conventional non-contrast computed tomography (CT), whereas their characteristic differentiating features are largely demonstrated on contrast-enhanced imaging. Lesion characterisation therefore remains challenging when iodinated contrast administration is limited, such as in patients with renal dysfunction or a history of hypersensitivity to iodinated contrast media, or when liver lesions are incidentally detected on non-contrast CT. This study aimed to develop and internally validate a diagnostic prediction model combining non-contrast dual-layer spectral detector CT (SDCT) quantitative parameters with clinical and laboratory indicators to differentiate hepatic hemangiomas (HH) from malignant liver tumours. Methods: This retrospective study included patients who underwent abdominal SDCT between January and June 2025. Eligibility required pretreatment non-contrast and contrast-enhanced spectral CT and a maximum lesion diameter >1 cm; liver lesions other than HH, HCC, or MLC were excluded. Diagnosis was based on surgical or biopsy pathology when available and otherwise on multimodal imaging findings combined with prior imaging. Eighteen predictors were evaluated: six clinical/laboratory variables and 12 non-contrast SDCT parameters. Variables with P<0.05 in univariable analysis underwent least absolute shrinkage and selection operator (LASSO) logistic regression, collinearity assessment, and multivariable logistic regression. Model performance was assessed using receiver operating characteristic (ROC) curves and area under the curve (AUC), DeLong testing, calibration curves, Brier scores, decision curve analysis (DCA), and 1,000 bootstrap resamples for internal validation. Results: A total of 119 patients were included: 63 with HH and 56 with malignant liver tumours, including 25 with HCC and 31 with MLC. Lymphocyte-to-monocyte ratio, the CT attenuation range on 70-keV virtual monoenergetic images, and electron density were retained in the combined model. The AUCs of the clinical, imaging, and combined models were 0.752 [95% confidence interval (CI): 0.665-0.840], 0.797 (95% CI: 0.715-0.879), and 0.866 (95% CI: 0.800-0.932), respectively. The combined model outperformed the clinical and imaging models on DeLong testing (P=0.002 and P=0.007, respectively). Its sensitivity, specificity, and accuracy were 75.0%, 85.7%, and 80.7%, respectively. Brier scores were 0.199, 0.178, and 0.146 for the clinical, imaging, and combined models, respectively. DCA suggested potential clinical net benefit, and the bootstrap optimism-corrected AUC of the combined model was 0.833. Conclusions: A diagnostic prediction model combining non-contrast SDCT quantitative parameters with clinical and laboratory indicators showed better discrimination than either the clinical or imaging model alone and acceptable calibration for the preliminary differentiation of HHs from malignant liver tumours. When contrast-enhanced imaging is limited, the model may support adjunctive risk stratification. Because only internal validation was performed, external validation in independent populations is required.
Source: Translational Cancer Research
Language: English
Subfield: Hepatology
Field: Medicine
Domain: Health Sciences
Open Access status: diamond
OpenAlex ID: W4416167661