A comparison of the diagnostic accuracy and reliability of subjective grading and computer-aided assessment of intranodal vascularity in differentiating metastatic and reactive cervical lymphadenopathy

Abstract

Purpose: Ultrasound is a well-established imaging modality in the assessment of malignant cervical lymphadenopathy. With the use of Doppler ultrasound, intranodal vascularity can be evaluated. However, the major limitation of ultrasound is operator dependency. Therefore, this study aimed to evaluate and compare the diagnostic accuracy and reliability of the subjective grading and computer-aided approach in assessing intranodal vascularity for the differentiation of benign and malignant lymph nodes. Materials and Methods: The present study retrospectively assessed 99 power Doppler ultrasound images of cervical lymph nodes and evaluated the degree of intranodal vascularity using qualitative subjective grading (QSG) and quantitative computer-aided (QCA) methods. The diagnostic accuracy of the two methods in distinguishing metastatic and reactive nodes and their inter- and intra-rater reliability in assessing intranodal vascularity were evaluated and compared. Results: The results showed that the QCA method was more accurate than the QSG method with a significantly higher sensitivity (67.8 % and 61.9 %, respectively, p < 0.05) and specificity (73.3 % and 57.3 %, respectively, p < 0.05). Using the intranodal vascularity index as determined by the QCA approach, the optimum cut-off to differentiate metastatic and reactive cervical lymph nodes was 32 %. The QCA method showed higher inter- and intra-rater reliability than the QSG method. Conclusion: In the assessment of the degree of intranodal vascularity, the QCA method was more accurate and reliable than the QSG method in distinguishing metastatic and reactive lymph nodes.