Liu Lingke, Hu Minfei, Zhou Yun, Zheng Fei, Ma Xiaohui, Yang Li, Xu Yiping, Teng Liping, Hu Bin, Lu Meiping, Xu Xuefeng
Department of Rheumatology Immunology & Allergy, The Children's Hospital, National Clinical Research Center for Child Health, Zhejiang University School of Medicine, Binsheng Rd 3333, Binjiang District, Hangzhou, 310052, P.R. China.
Department of Pediatrics, The Affiliated Hospital of Shaoxing University, Shaoxing, 312000, PR China.
BMC Pediatr. 2025 Aug 2;25(1):589. doi: 10.1186/s12887-025-05968-z.
Interstitial lung disease (ILD) is a frequently observed pulmonary manifestation in juvenile dermatomyositis (JDM), and may significantly impact patient outcomes. Therefore, early lung involvement identification is essential. Radiomics is a new image analysis technique and might offer valuable information for the diagnosis of interstitial lung disease in juvenile dermatomyositis (JDM-ILD).
We retrospectively analyzed clinical data of 56 children with JDM, and all participants gave written informed consent. These children were divided into the JDM group (n = 32) and JDM-ILD group (n = 24) based on chest high-resolution CT (HRCT). The lung intelligence kit (LK) software was used to outline the bilateral lung tissue structure automatically. The radiomics score combining with clinical variables was used to establish a prediction model for JDM-ILD.
A total of seven radiomics features including the maximum, mean, skewness, and kurtosis features for the First Order Features, the InverseVariance feature for the Gray Level Co-occurrence Matrix (GLCM) Features, the Size Zone NonUniformity Normalized feature for the Gray Level Size Zone Matrix(GLSZM)Features, and the Run Entropy feature for the Gray Level Run Length Matrix (GLRLM) Features were identified. The multivariable logistic regression revealed that anti-MDA5 antibody and radiomics score showed a significant correlation with the development of ILD in children with JDM. The combined prediction model based on radiomics score and anti-MDA5 antibody achieved good performance in predicting JDM-ILD in the training (0.92, 95% CI 0.82-1.00) and validation (0.93, 95% CI 0.83-1.00) groups.
The nomogram combining radiomics and clinical variables achieved an optimal prediction of ILD in children with JDM. This approach may facilitate earlier detection of pulmonary involvement, support individualized risk stratification, and guide more proactive therapeutic interventions. Future studies should aim to validate this model in larger, prospective, and multicenter cohorts.
间质性肺疾病（ILD）是幼年皮肌炎（JDM）中常见的肺部表现，可能会显著影响患者预后。因此，早期识别肺部受累情况至关重要。放射组学是一种新的图像分析技术，可能为幼年皮肌炎相关性间质性肺疾病（JDM-ILD）的诊断提供有价值的信息。
我们回顾性分析了56例JDM患儿的临床资料，所有参与者均签署了书面知情同意书。根据胸部高分辨率CT（HRCT）将这些患儿分为JDM组（n = 32）和JDM-ILD组（n = 24）。使用肺部智能套件（LK）软件自动勾勒双侧肺组织结构。将放射组学评分与临床变量相结合，建立JDM-ILD的预测模型。
共识别出7个放射组学特征，包括一阶特征中的最大值、均值、偏度和峰度特征，灰度共生矩阵（GLCM）特征中的逆方差特征，灰度大小区域矩阵（GLSZM）特征中的大小区域非均匀性归一化特征，以及灰度游程长度矩阵（GLRLM）特征中的游程熵特征。多变量逻辑回归显示，抗MDA5抗体和放射组学评分与JDM患儿ILD的发生显著相关。基于放射组学评分和抗MDA5抗体的联合预测模型在训练组（0.92，95%CI 0.82 - 1.00）和验证组（0.93，95%CI 0.83 - 1.00）中对JDM-ILD的预测表现良好。
结合放射组学和临床变量的列线图对JDM患儿的ILD进行了最佳预测。这种方法可能有助于更早地检测肺部受累情况，支持个体化风险分层，并指导更积极的治疗干预。未来的研究应旨在在更大规模、前瞻性和多中心队列中验证该模型。
