Is Computer-Aided Quantification of Intranodal Vascularity a Useful Tool for Differentiating Different Causes of Cervical Lymphadenopathy

Abstract

Advancement of information technology allows accurate and reliable assessment of medical images including ultrasound images. This study investigated the feasibility of using intranodal vascularity index (VI) as evaluated by computer-aided quantification in differentiating the various common causes of cervical lymphadenopathy. Power Doppler sonograms of 347 patients with palpable cervical lymph nodes were reviewed (155 metastases, 23 lymphoma, 44 tuberculous lymphadenitis, 125 reactive). Ultrasound images were evaluated and VI of lymph nodes was quantified using our self-developed computer algorithm. The diagnostic performance of using intranodal VI in distinguishing different disease groups was evaluated and compared. Metastatic and lymphomatous lymph nodes tend to be more vascular than tuberculous and reactive nodes. Intranodal VI had the highest diagnostic accuracy in distinguishing metastatic and tuberculous nodes with a sensitivity of 80%, specificity of 73%, positive predictive value of 91%, negative predictive value of 51% and overall accuracy of 68% when using a cut-off VI of 22%. Computer-aided quantification allows objective assessment of intranodal vascularity. Intranodal VI is a useful parameter in differentiating metastatic and tuberculous lymph nodes, but has limited value in distinguishing lymphomatous nodes from metastatic and reactive nodes.