Efficacy of logistic regression model based on multiparametric ultrasound in assessment of cervical lymphadenopathy - a retrospective study

Abstract

Objectives: To investigate whether a multiparametric ultrasound (MPUS) diagnostic model improves differential diagnosis of benign and malignant cervical lymph nodes. Methods: MPUS evaluation was performed on 87 lesions in 86 patients, and related characteristics and parameters of the patients and lesions were studied and logistic regression models based on the MPUS characteristics of cervical lymph nodes were built. A receiver operating characteristic curve and area under the curve (AUC) were built for the evaluation of diagnostic performances. Results: Of the 87 lesions in 86 patients, there were 31 benign and 56 malignant lesions. Regression models for Duplex ultrasound and MPUS were established. The Duplex ultrasound regression model showed a sensitivity, specificity, positive predictive value and negative predictive value of 94.4, 61.3, 86.3 and 80.9%, respectively. The predictive accuracy was 82.4%, and the AUC was 0.861. The MPUS regression model showed a sensitivity, specificity, positive predictive value and negative predictive value of 98.1, 61.3, 81.5 and 95.0%, respectively. The predictive accuracy was 84.7%, and the AUC was 0.894. The differences in AUCs between the Duplex ultrasound model and MPUS model, ultrasound model and ultrasonic elastography (UE), and Duplex ultrasound and UE were not significant (all p > 0.05); the differences in AUCs between the MPUS model and Duplex ultrasound, Duplex ultrasound model and Duplex ultrasound, and MPUS model and UE were significant (all p < 0.05). Conclusions: The Duplex ultrasound and MPUS models achieve significantly higher diagnostic performance for differentiating between benign and malignant cervical lymph nodes.