Quantification of intranodal vascularity by computer pixel-counting method enhances the accuracy of ultrasound in distinguishing metastatic and tuberculous cervical lymph nodes

Sammy C. H. Cheng, Anil T. Ahuja, Michael Ying

Abstract

Background: Ultrasound is a common imaging method for assessment of cervical lymph nodes. However, metastatic and tuberculous lymph nodes have similar sonographic features in routine ultrasound examination. Computer-aided assessment could be a potential adjunct to enhance the accuracy of differential diagnosis.
Methods: Gray-scale and power Doppler sonograms of 100 patients with palpable cervical lymph nodes were reviewed and analyzed (60 metastatic nodes, 40 tuberculous nodes). Final diagnosis of lymph nodes was based on fine needle aspiration and cytology. Sonograms were reviewed and assessed for nodal shape, echogenic hilus, intranodal necrosis and vascular distribution (conventional assessment). Intranodal vascularity was quantified using a customized computer algorithm to determine vascularity index (VI). The diagnostic accuracy of using conventional assessment and its combination with intranodal VI method was evaluated and compared.
Results: Metastatic and tuberculous nodes tended to be round (75.0% vs. 50.0%), without echogenic hilus (86.7% vs. 72.5%) and have peripheral vascularity (73.3% vs. 85.0%). Intranodal necrosis is more common in tuberculous nodes (27.5%) than metastatic nodes (8.3%). Using conventional assessment in differentiating metastatic and tuberculous nodes, the diagnostic accuracy was 56% with a sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of 56.7%, 55%, 65.4% and 45.8% respectively. The VI of metastatic nodes (23.4%±2.1%) was significantly higher than that of tuberculous nodes (12.0%±1.6%) (P<0.05). The optimum cut-off of VI for the differential diagnosis was 20%. By combining conventional assessment and intranodal VI quantification, the diagnostic accuracy was increased to 69% with a sensitivity, specificity, PPV and NPV of 80%, 52.5%, 71.6%, 63.6% respectively. The increase in sensitivity was statistically significant (P=0.006).
Conclusions: Computer-aided quantification of intranodal vascularity provides added value in routine ultrasound assessment of cervical lymph nodes. It enhances the accuracy of ultrasound in distinguishing metastatic and tuberculous cervical lymph nodes.