基于MRI影像组学预测腰椎间盘突出症针刺康复治疗临床疗效的初步探讨  

Preliminary exploration of predicting clinical efficacy after acupuncture and rehabilitation therapy for lumbar disc herniation based on Radiomics Features of MRI

作　　者：刘荣 肖利 蒯林 武朦 韩超 LIU Rong;XIAO Li;KUAI Lin;WU Meng;HAN Chao(Department of Radiology,Pidu Hospital of Traditional Chinese Medcine in Chengdu City,Chengdu 611730,China)

机构地区：[1]成都市郫都区中医医院医学影像科,成都611730

出　　处：《磁共振成像》2025年第4期93-98,共6页Chinese Journal of Magnetic Resonance Imaging

基　　金：成都中医药大学“杏林学者”学科人才科研提升计划项目(编号:XJ2023022901)。

摘　　要：目的 初步探讨基于MRI横轴位T2WI构建的影像组学模型用于预测腰椎间盘突出症(lumbar disc herniation,LDH)针刺康复治疗临床疗效的价值。材料与方法 回顾性分析155例经住院完成1个疗程针刺康复治疗的LDH患者资料(治疗有效102例,治疗无效53例),在R软件中,用Sample函数将患者按照7∶3的比例分为训练集(n=108)与测试集(n=47)。采用3D-Slicer软件对每例患者基于治疗前磁共振T2WI横轴位图像提取851个影像组学特征,运用Pearson或Spearman相关分析删除冗余特征后,用最小绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)回归分析对特征进一步降维与模型构建,在训练集与测试集中应用受试者工作特征(receiver operating characteristic,ROC)曲线对模型进行验证,评价影像组学模型预测LDH针刺康复治疗临床疗效的效能。结果 经筛选后7个影像组学特征用于构建LDH针刺康复治疗临床疗效预测模型。训练集中预测模型的ROC曲线下面积(area under the curve,AUC)为0.862(95%置信区间:0.789~0.935),敏感度82.5%,特异度75.0%;测试集中AUC为0.887(95%置信区间:0.773~0.989),敏感度92.3%,特异度79.4%。结论 基于MRI横轴位T2WI影像组学模型在LDH针刺康复治疗临床疗效预测中有较好的性能,可作为临床决策的有力依据。Objective:To explore the value of an radiomics model based on MRI transverse axial plain scan in predicting the clinical efficacy of acupuncture and moxibustion rehabilitation therapy for lumbar disc herniation(LDH).Materials and Methods:A retrospective analysis was conducted on 155 LDH patients who had completed one courses of acupuncture and moxibustion rehabilitation treatment in hospital(102 cases were effective and 53 cases were ineffective).In R software,the patients were divided into a training set(n = 108) and a testing set(n = 47) using the sample function at a ratio of 7∶3.For each patient,851 radiomics features were extracted using 3D-Slicer software based on the pre-treatment magnetic resonance T2WI transverse axial image.After removing redundant features using Pearson or Spearman correlation analysis,the least absolute shrinkage and selection operator(LASSO)regression analysis was used to further reduce the dimensionality of the features and construct the model.The receiver operating characteristic(ROC) curve was applied to the training and testing sets to validate the model and evaluate the efficacy of radiomics features in predicting the clinical efficacy of acupuncture rehabilitation therapy for LDH.Results:After feature selection,7 radiomics features were used to construct a clinical efficacy prediction model for LDH acupuncture and moxibustion rehabilitation therapy.The area under the ROC curve(AUC) of the prediction model in the training set was 0.862 [95% confidence interval(CI):0.789 to 0.935],with sensitivity and specificity of 82.5% and 75.0%,respectively;the AUC of the test set was 0.887(95% CI:0.773 to 0.989),with sensitivity and specificity of 92.3% and 79.4%,respectively.Conclusions:Based on the MRI axial imaging-based radiomics model,it exhibits excellent predictive performance for the clinical efficacy of acupuncture and rehabilitation therapy for LDH,serving as a robust basis for clinical decision-making.

关 键 词：腰椎间盘突出症 针刺 康复治疗 磁共振成像 影像组学 疗效预测 

分 类 号：R445.2[医药卫生—影像医学与核医学] R681.53[医药卫生—诊断学]