Department of Radiology, Seoul National University Bundang Hospital, 82 Gumi-ro 173 Beon-gil, Bundang-gu, Seongnam, 13620, South Korea.
Department of Radiology, Seoul National University College of Medicine, 103 Daehak-ro, Jongno-gu, Seoul, 03080, South Korea.
Eur Radiol. 2020 Aug;30(8):4475-4485. doi: 10.1007/s00330-020-06794-w. Epub 2020 Mar 18.
To establish a diagnostic tree analysis (DTA) model based on ultrasonography (US) findings and clinical characteristics for differential diagnosis of common causes of cervical lymphadenopathy in children.
A total of 242 patients (131 boys, 111 girls; mean age, 11.2 ± 0.3 years; range, 1 month-18 years) with pathologically confirmed Kikuchi disease (n = 127), reactive hyperplasia (n = 64), lymphoma (n = 24), or suppurative lymphadenitis (n = 27) who underwent neck US were included. US images were retrospectively reviewed to assess lymph node (LN) characteristics, and clinical information was collected from patient records. DTA models were created using a classification and regression tree algorithm on the basis of US imaging and clinical findings. The patients were randomly divided into training (70%, 170/242) and validation (30%, 72/242) datasets to assess the diagnostic performance of the DTA models.
In the DTA model based on all predictors, perinodal fat hyperechogenicity, LN echogenicity, and short diameter of the largest LN were significant predictors for differential diagnosis of cervical lymphadenopathy (overall accuracy, 85.3% and 83.3% in the training and validation datasets). In the model based on categorical parameters alone, perinodal fat hyperechogenicity, LN echogenicity, and loss of fatty hilum were significant predictors (overall accuracy, 84.7% and 86.1% in the training and validation datasets).
Perinodal fat hyperechogenicity, heterogeneous echotexture, short diameter of the largest LN, and loss of fatty hilum were significant US findings in the DTA for differential diagnosis of cervical lymphadenopathy in children.
• Diagnostic tree analysis model based on ultrasonography and clinical findings would be helpful in differential diagnosis of pediatric cervical lymphadenopathy. • Significant predictors were perinodal fat hyperechogenicity, heterogeneous echotexture, short diameter of the largest LN, and loss of fatty hilum.
建立基于超声（US）表现和临床特征的诊断树分析（DTA）模型，以鉴别儿童颈部淋巴结肿大的常见病因。
本研究共纳入 242 例经病理证实为 Kikuchi 病（n=127）、反应性增生（n=64）、淋巴瘤（n=24）或化脓性淋巴结炎（n=27）的患儿，所有患儿均接受了颈部 US 检查。回顾性分析 US 图像以评估淋巴结（LN）特征，并从病历中收集临床信息。基于 US 成像和临床发现，使用分类和回归树算法建立 DTA 模型。将患者随机分为训练集（70%，170/242）和验证集（30%，72/242），以评估 DTA 模型的诊断性能。
在基于所有预测因素的 DTA 模型中，LN 周围脂肪高回声、LN 回声强度和最大 LN 短径是鉴别颈部淋巴结肿大的重要预测因素（在训练和验证数据集中，整体准确性分别为 85.3%和 83.3%）。在基于分类参数的模型中，LN 周围脂肪高回声、LN 回声强度和脂肪门消失是重要的预测因素（在训练和验证数据集中，整体准确性分别为 84.7%和 86.1%）。
LN 周围脂肪高回声、异质性回声、最大 LN 短径和脂肪门消失是儿童颈部淋巴结肿大 DTA 鉴别诊断的重要 US 表现。
• 基于超声和临床特征的诊断树分析模型有助于鉴别儿童颈部淋巴结肿大。
• 重要的预测因素包括 LN 周围脂肪高回声、异质性回声、最大 LN 短径和脂肪门消失。
