Journal of Medical Ultrasound (Jan 2019)
Comparison of strain elastography, shear wave elastography, and conventional ultrasound in diagnosing thyroid nodules
Abstract
Objective: The purpose of this study is to compare the diagnostic performances of strain elastography (SE), shear wave elastography (SWE), and traditional ultrasound (US) features in diagnosing thyroid nodules. Subjects and Methods: This study included 185 adult patients with thyroid nodules who underwent conventional gray-scale US, SE, and SWE. SE was scored using a four-pattern elastographic scoring (ES) system. SWE values were presented as mean SWE values and standard derivation using Young's modules. The optimal cutoff values of the mean SWE values for predicting malignancy were determined using receiver operating characteristic (ROC) curve analysis. We used logistic regression models to test elastography as a novel significant predictor for the diagnosis of malignant nodules. The diagnostic performance of elastography parameters was compared with a traditional trained model. Results: Malignant thyroid nodules were stiffer for SE (ES patterns 1 and 2/3 and 4) and mean SWE values (4/17; 51.0 ± 24.4 kPa) than for benign nodules (114/50; 33.1 ± 25.2 kPa) (P < 0.01). In ROC curve analyses, a mean SWE value of 32 kPa was the optimal cutoff point, with diagnostic performance measures of 81% sensitivity, 65% specificity, a 23% positive predictive value (PPV), and 96% negative predictive value (NPV). In multivariate logistic regression, the mean SWE value (≥32 kPa) was an independent predictor for malignancy (odds ratio: 16.8; 95% confidence interval [CI]: 3.6–78.3). However, after the addition of SE and SWE to traditional US features, the C-statistic was not significantly increased compared to the traditional model (0.88, 95% CI: 0.81–0.94 vs. 0.91, 0.85–0.97, P = 0.4). Conclusion: In this study, we confirmed SWE as an independent predictor for malignant thyroid nodules. However, in comparing the new extended elastography model to our previous prediction model, the new extended model showed no significant difference in the diagnostic performance.
Keywords