摘要
目的:本研究旨在探讨中老年男性骨质疏松症的MRI影像组学模型构建与应用研究。方法:收集2023年1月-2024年1月于韶关市第一人民医院就诊的中老年男性患者128例作为研究对象,通过同期行腰椎MRI及双能X射线骨密度测量,将骨量正常组为对照,以双能X射线骨密度检查结果为诊断金标准进行分析,选择独立的MRI序列与T1WI+T2WI联合序列,通过美国AK软件中的Pyradiomics模块提取影像组学特征,Spearman分析特征集合,假如结果中依旧存在着冗余表现,再次使用梯度提升决定树(GBDT)算法进行降维。分析单独序列影像组学模型ROC曲线参数与联合影像组学和影像组学模型诊断性能参数。结果:通过T1WI和T2WI序列的腰椎MRI平扫图像提取影像组学特征,并建立逻辑回归模型(包括T1WI模型、T2WI模型和T1WI+T2WI联合模型)。128例患者中包含55例骨质正常或减低者,73例骨质疏松症者。T1WI序列AUC为0.801,95%置信区间为0.700~0.910、敏感性为0.845、准确性为0.719、特异性为0.6245;而T2WI AUC为0.818,95%置信区间为0.712~0.917,敏感性为0.769、准确性为0.738、特异性为0.836;T1WI+T2WI的敏感性为77.15,特异性为100.00,准确性为84.0,阳性预测值为100,阴性预测值为84.78;影像组学模型敏感性为90.88,特异性为96.31,准确性为92.0,阳性预测值为95.18,阴性预测值为93.08,且按照Radscore的截断值可以将患者分为高风险组与低风险组。与单独序列模型相较来说,联合模型的敏感性、特异性等均比较高;影像组学模型的敏感性、特异性、准确性高于双能X射线骨密度测量(P<0.05)。结论:在中老年男性腰椎骨质疏松的预测中,基于MRI平扫的影像组学模型具有较高的诊断效能,是较为有效的辅助工具,可以为临床决策以及患者预后的改善提供帮助。
Objective:The purpose of this study was to investigate the construction and application of MRI imaging omics model for osteoporosis in middle-aged and elderly men.Methods:We collected 128 middle-aged and elderly male patients who were treated in the First People's Hospital of Shaoguan City from January 2023 to January 2024 as the study objects.Lumbar MRI and dual-energy X-ray bone mineral density measurement were performed during the same period.The normal bone mass group was taken as the control group,and the results of dual-energy X-ray bone mineral density examination were used as the diagnostic gold standard for analysis.The independent MRI sequence and the joint sequence of T1WI+T2WI were selected,and the image omics features were extracted by Pyradiomics module in American AK software.Spearman's correlation analyzed the feature set.If there was still redundancy in the results,gradient boosting decision tree(GBDT)algorithm was used again for dimensionality reduction.ROC curve parameters of single sequence imaging model and diagnostic performance parameters of combined imaging model and imaging model were analyzed.Results:The imaging features were extracted and logistic regression models(including T1WI model,T2WI model and T1WI+T2WI combined model)were established based on T1WI and T2WI plain MRI images.Among the 128 patients,55 had normal or reduced bone mass and 73 had osteoporosis.The AUC of T1WI sequence was 0.801,the 95%confidence interval was 0.700-0.910,the sensitivity was 0.845,the accuracy was 0.719,and the specificity was 0.6245.The AUC of T2WI was 0.818,95%confidence interval was 0.712-0.917,sensitivity was 0.769,accuracy was 0.738,specificity was 0.836.The sensitivity of T1WI+T2WI was 77.15,specificity was 100.00,accuracy was 84.0,positive predictive value was 100,and negative predictive value was 84.78.The sensitivity of the imaging omics model was 90.88,the specificity was 96.31,the accuracy was 92.0,the positive predictive value was 95.18,the negative predictive value was 93.08,and the patients could be divided into high risk group and low risk group according to the truncation value of Radscore.Compared with the single sequence model,the sensitivity and specificity of the combined model are higher.The sensitivity,specificity and accuracy of the imaging omics model were higher than those of dual-energy X-ray bone density measurement(P<0.05).Conclusion:In the prediction of lumbar osteoporosis in middle-aged and elderly men,the imaging omics model based on MRI plain scan has high diagnostic efficacy and is a more effective auxiliary tool,which can help clinical decision-making and improve the prognosis of patients.
作者
胡运祥
朱艳敏
谢小清
熊伟坚
HU Yunxiang;ZHU Yanmin;XIE Xiaoqing;XIONG Weijian(Medical Imaging Department,Shaoguan First People's Hospital,Guangdong 512000,China)
出处
《影像技术》
2025年第1期8-13,23,共7页
Image Technology
基金
韶关市卫生健康科研项目(Y24107)。
关键词
影像组学模型
骨质疏松
腰椎MRI平扫
中老年男性
image omics model
osteoporosis
lumbar MRI plain scan
middle-aged and elderly men
