- 期刊
- OpenAccess
Ultrasound Texture Analysis: The Association With Neck Metastatic Lymphadenopathy
摘要
Background: High-resolution ultrasound (US) is widely used in neck lymph node assessment. Texture analysis has been applied to US image analysis and has potential in US differentiation. The aims of this study were to perform US texture analysis for neck lymphadenopathy (LAP) and validate a previous US prediction model. Methods: We collected 219 patients with neck LAP, including 103 female and 116 male patients. We performed texture analysis with 2D-grayscale images for benign and malignant neck LAP and validated our previous US prediction model. Results: Entropy was significantly higher at 0, 45, 90, and 135 degrees, as was the mean gray level cooccurrence in texture analysis for metastatic carcinoma. There were no statistically significant differences for other texture characteristics, including angular second moment, contrast, correlation, inverse difference moment, dissimilarity, inverse difference variance, cluster shade, and cluster prominence. Entropy (mean) was a significant predictor of univariate logistic analysis (OR: 2.35, 95% CI: 1.44-3.81); however, entropy was not an independent factor in multivariate analysis (OR: 1.58, 95% CI: 0.92-2.68) after adjusting for age, sex and S/ L ratio. The sensitivity, specificity, PPV, NPV and overall accuracy were 87.3 (80.0-95.1)%, 78.4 (71.7-85.0)%, 66.0 (56.4-75.5)%, 92.8 (88.3-97.3)% and 81.3 (76.1-86.4), respectively, for the US prediction model. Conclusion: Entropy is a significant differentiating factor for metastatic neck LAP and is associated with internal echogenicity. However, entropy alone was not an independent factor in predicting metastatic LAP.