Objectives: To establish a diagnostic tree analysis (DTA) model based on computed tomography (CT) findings and clinical information for differential diagnosis of cervical necrotic lymphadenopathy, especially in regions where tuberculous lymphadenitis and Kikuchi disease are common. Methods: A total of 290 patients (147 men and 143 women; mean age (years), 46.2 ± 19.5; range, 3–91) with pathologically confirmed metastasis (n = 110), tuberculous lymphadenitis (n = 73), Kikuchi disease (n = 71), and lymphoma (n = 36) who underwent contrast-enhanced neck CT were included. The patients were randomly divided into training (86%, 248/290) and validation (14%, 42/290) datasets to assess diagnostic performance of the DTA model. Two sorts of DTA models were created using a classification and regression tree algorithm on the basis of CT findings alone and that combined with clinical findings. Results: In the DTA model based on CT findings alone, perinodal infiltration, number of the necrotic foci, percentage of necrotic lymph node (LN), degree of necrosis, margin and shape of the necrotic portion, shape of the LN, and enhancement ratio (cutoff value, 1.93) were significant predictors for differential diagnosis of cervical necrotic lymphadenopathy. The overall accuracy was 80.6% and 73.8% in training and validation datasets. In the model based on imaging and clinical findings, tenderness, history of underlying malignancy, percentage of necrotic LN, degree of necrosis, and number of necrotic foci were significant predictors. The overall accuracy was 87.1% and 88.1% in training and external validation datasets. Conclusions: The DTA model based on CT imaging and clinical findings may be helpful for the diagnosis of cervical necrotic lymphadenopathy. Key Points: • The diagnostic tree analysis model based on CT may be useful for differential diagnosis of cervical necrotic lymphadenopathy. • Perinodal infiltration, number of necrotic foci, percentage of necrotic lymph nodes, degree of necrosis, margin and shape of necrotic portion, lymph node shape, and enhancement ratio were the most significant predictors.
ASJC Scopus subject areas
- Radiology Nuclear Medicine and imaging