基于超声影像组学对痛风性关节炎急性发作的预测及与炎症因子水平的关系OACSTPCD
Prediction of Acute Attack of Gouty Arthritis Based on Ultrasound Radiomics and Its Relationship with the Levels of Inflammatory Factors
目的:探讨基于超声影像组学对痛风性关节炎急性发作的预测,并分析其与炎症因子水平的关系.方法:回顾性收集2022年7月至2023年10月于本院就诊的138名痛风性关节炎患者的超声影像学数据和血清炎症因子水平,采用计算机产生随机数法以7:3比例分为训练集(96例)和测试集(4 2例),并根据是否急性发作将训练集患者分为急性发作组(3 9例)和非急性发作组(5 7例).对比两组患者一般资料,通过影像组学方法,从超声图像中提取特征,进一步利用机器学习算法建立预测模型.并利用受试者操作特性(ROC)曲线、校准曲线以及临床决策曲线,对模型的预测效能进行评估和衡量.同时检测患者炎症因子水平.通过相关性分析,探讨炎症因子水平与超声影像特征的关系.结果:单因素分析表明,年龄、总胆红素、血糖、高尿酸水平、γ-干扰素(IFN-γ)、转化生长因子β(TGF-β)、肌酐、肿瘤坏死因子α(TNF-α)、血尿素氮、三酰甘油、总胆固醇、血小板计数、超声表现及超声评分方面具有显著差异性,差异有统计学意义(P＜0.05);多因素分析结果表明,年龄、总胆红素、血糖、高尿酸水平、TGF-β、IFN-γ、TNF-α、超声表现(双轨征、滑膜炎)、超声评分(滑膜病损、滑膜内血流信号、骨侵蚀、关节囊物质沉积)均是痛风性关节炎患者急性发作的影响因素(P＜0.05).筛选出影像组学标签5个特征(ICC及95％CI均＞0.800),在未急性发作组和急性发作组中差异性显著(P＜0.05).训练集和测试集模型的ROC曲线、校准曲线及临床决策曲线提示该模型区分度及临床净收益良好.Pearson相关性分析显示,3个炎症因子(TGF-β、IFN-γ和TNF-α)与45项特征呈极强相关,与33项特征呈强相关.结论:超声影像组学可以作为预测痛风性关节炎急性发作的有效工具,并且与炎症因子水平密切相关.
Objective:To explore the prediction of acute attack of gouty arthritis based on ultrasound radiomics,and analyze its relationship with the levels of inflammatory factors.Methods:The ultrasonic imaging data and serum inflammatory factor levels of 138 patients with gouty arthritis who were treated in our hospital from July 2022 to October 2023 were retrospectively collected and divided into the training set(96 cases)and the test set(42 cases)with a ratio of 7:3 by using computer-generated random number method.The patients in the training set were divided into acute attack group(39 cases)and non-acute attack group(57 cases)according to whether they had acute attack or not.By comparing the general data of the two groups of patients,the features were extracted from the ultrasound images by means of radiomics methods,and the prediction model was further established by machine learning algorithm.Receiver operating characteristic(ROC)curve,calibration curve and clinical decision curve were used to evaluate and measure the predictive efficacy of the model.At the same time,the levels of inflammatory factors were detected.Through correlation analysis,the relationship between the levels of inflammatory factors and ultrasonic image features was discussed.Results:Univariate analysis showed significant differences in age,total bilirubin,blood glucose,high uric acid level,interferon-γ(IFN-γ),transforming growth factor-β(TGF-β),creatinine,tumor necrosis factor-α(TNF-α),blood urea nitrogen,triglyceride,total cholesterol,platelet count,ultrasound findings and ultrasound scores,with statistical significance(P＜0.05).The results of multivariate analysis showed that age,total bilirubin,blood glucose,high uric acid level,TGF-β,IFN-γ,TNF-α,ultrasound manifestations(double track sign,synovitis),ultrasound scores(synovial lesions,synovial blood flow signals,bone erosion,joint capsule material deposition)were the influencing factors of acute attack in patients with gouty arthritis(P＜0.05).Five radiomic features were identified(ICC and 95％CI＞0.800),without acute group and the differences in acute group significantly(P＜0.05).ROC curves,calibration curves and decision curves of training set and test set models indicate that the model has good differentiation and clinical net benefit.Pearson correlation analysis showed that three inflammatory factors,TGF-β,IFN-y,and TNF-α,were strongly associated with 45 features and 33 features.Conclusion:Ultrasound radiomics can be used as an effective tool to predict the acute attack of gouty arthritis,and it is closely related to the levels of inflammatory factors.
朱晓霞;鲁健;施亚男
启东市人民医院超声科,江苏南通 226200启东市人民医院内分泌科,江苏南通 226200
临床医学
痛风性关节炎超声影像组学急性发作预测炎症因子
gouty arthritisultrasound radiomicsacute attackpredictioninflammatory factor
《影像科学与光化学》 2024 (005)
459-470 / 12
2022年度南通市基础科学研究和社会民生科技计划项目(MSZ2022061)
评论