非增强MRI、静态和动态对比增强MRI预测脊柱压缩性骨折良、恶性的Logistic回归分析

Logistic regression analysis of non-enhanced MR imaging,static and dynamic enhanced MR imaging in prediction of malignancy of vertebral compression fractures

目的:评价非增强磁共振成像(Magnetic resonance imaging,MRI)、静态增强MRI和动态对比增强MRI对脊柱良、恶性压缩性骨折的鉴别诊断价值并确定不同检查方法中预测脊柱压缩性骨折良、恶性的最主要的MRI征象。方法:经临床和(或)病理组织证实的胸、腰椎体骨质疏松伴椎体急性压缩性骨折(Acute comtression fracture,ACF)43例(71个椎体)和胸、腰椎体恶性压缩性骨折(Malignant comtression fracture,MCF)49例(99个椎体)行非增强、钆剂T1WI静态和动态对比增强MR检查。以脊柱压缩性骨折的良、恶性作为因变量,各种MRI征象为自变量因素分别采用非增强MR(I模型1)、非增强MRI联合静态增强MRI(模型2)以及非增强MRI、静态增强MRI征象联合动态对比增强MR(I模型3)对脊柱压缩性骨折良、恶性的预测效果进行二值多元Logistic回归分析。不同的模型对脊柱压缩性骨折良、恶性预测的敏感性、特异性及准确性采用受试者操作特性曲线(Receiver operating characteristic,ROC)分析。结果:单因素分析结果显示:椎体T1WI信号均匀性、椎体后缘膨隆、椎体后上角后凸、椎弓根改变、硬膜外软组织肿块、椎旁软组织改变、椎体强化模式、椎旁软组织强化模式、椎体动态增强开始时间、时间信号强度曲线(Time-intensity curve,TIC)类型共10个征象在脊柱良、恶性压缩性骨折组间的差异有统计学意义(P值均<0.05),椎体T2WI信号均匀性、椎体内液体征象、压缩椎体形态、椎间盘改变共4个征象在脊柱良、恶性压缩性骨折组间的差异无统计学意义(P值均>0.05)。ROC分析结果显示:非增强MRI、静态增强MRI联合动态对比增强MRI预测脊柱压缩性骨折良、恶性的效果优于非增强MRI或非增强MRI联合静态增强MRI,其诊断的敏感性、特异性和准确性分别为81.6%、90.7%和85.9%。Logistic回归分析结果显示:椎弓根改变、椎体强化模式以及TIC类型是预测脊柱压缩性骨折良、恶性的最主要的影响因素(OR值分别为3.558,4.666和0.007,P值均<0.05)。结论:非增强MRI、静态增强MRI联合动态对比增强MRI可以提高脊柱良、恶性压缩性骨折的鉴别诊断能力。

Objective：To evaluate non-enhanced MR imaging,static and dynamic enhanced MR imaging in differentiation of benign from malignant vertebral compression fractures and to evaluate which MR imaging parameters are most predictive of malignancy.Methods：Non-enhanced MR imaging,static and dynamic gadolinium-enhanced MR imaging were performed on 43 patients with acute osteoporotic compression fracture and 49 patients with malignant compression fracture proved by clinical follow-up and/or histologic findings.Binary multivariate logistic regression analysis was used to identify the best combination of MR imaging parameters that might be predictive of malignancy.Subjective overall performance of diagnostic models for sensitivity,specificity and accuracy was evaluated with receiver operating characteristic curve（ROC）.Results：Univariate analysis showed that the frequency distribution of MR imaging parameters including T1 weighted imaging homogeneity of vertebrae,convex posterior of vertebrae,retropulsion of the posterosuperior angle of vertebrae,pedicle involvement,epidural soft tissue mass,paravertebral soft tissue lesions,pattern of enhancement of vertebrae,pattern of enhancement of paravertebral soft tissue lesions,start of dynamic enhancement and pattern of time-intensity curve（TIC） were different between benign and malignant compression fracture（P〈0.05）.However,T2 weighted imaging homogeneity of vertebrae,fluid sign,shape of vertebral collapse and intervertebral discinvolvement were not different between benign and malignant compression fracture（P〉0.05）.ROC analysis showed that combined non-enhanced,static and dynamic enhanced MR imaging parameters（model 3） were superior to non-enhanced MR imaging parameter（smodel 1）iction of malignancy.The sensitivity,specificity and accuracy of model 3 were 81.6%,90.7% and 85.9%,respectively.Multivariate logistic regression analysis revealed the most discriminating parameters were pedicle involvement,pattern of enhancement of vertebrae and pattern of TIC（OR value were 3.558,4.666 and 0.007,respectively,P〈0.05）.Conclusion：Non-enhanced MR imaging combined with static and dynamic enhanced MR imaging can improve the differentiation between benign and malignant vertebral compression fracture.