OBJECTIVES: The aim of this study was to establish prediction models for gallbladder lesions (GBLs) malignancy risk using contrast-enhanced ultrasound (CEUS) and arrival-time parametric imaging (ATPI). METHODS: From January 2013 to July 2020, a total of 59 patients in Peking University Cancer Hospital with GBLs were retrospectively enrolled in this study. All patients underwent ultrasound (US), CEUS, and subsequent ATPI analysis before receiving surgical treatment and getting a histopathological diagnosis. Univariate and multivariate analyses were used to select the independent predictors for malignancy. CEUS and CEUS + ATPI prediction models were established and displayed by nomogram. Receiver operating characteristic and calibration curves were drawn to evaluate the predictive ability of the models. Decision curve analysis was used to assess the clinical utility of the models. RESULTS: On univariate and multivariate analysis, diameter (p < .001), shape (p = .030), and washout time (p = .011) were included in the CEUS model, whereas diameter (p = .001), washout time (p = .012), and vascular pattern (p < .001) were incorporated into the CEUS + ATPI model. The AUC of the CEUS and CEUS + ATPI models were 0.908 and 0.978, respectively. Both models yielded a good net benefit for almost all threshold probabilities. CONCLUSION: Prediction models based on CEUS and ATPI can markedly enhance diagnostic accuracy of GBLs by providing information such as a more accurate size, shape, arterial vascular pattern, and washout time in a more intuitive and quantitative way.
Tang et al. (Wed,) studied this question.