Helmberger Thomas K, Laubenberger Jörg, Rummeny Ernst, Jung Gregor, Sievers Klaus, Döhring Wilfried, Meurer Karoline, Reiser Maximilian F
Institute of Clinical Radiology, Klinikum Grosshadern, Ludwig Maximilians University, Marchioninistrasse 15, 81377 Munich, Germany.
Eur Radiol. 2002 Jan;12(1):62-70. doi: 10.1007/s003300101122. Epub 2001 Oct 19.
The aim of this study was to determine if different types of focal hepatic lesions can be differentiated by specific quantitative and qualitative imaging characteristics pre- and post-Mangafodipir trisodium (MnDPDP) administration using a computerized multivariable, discriminant analysis (DA). In a multicenter trial, 151 patients with focal liver disease were studied at 1.5 and 1.0 T using gradient-recalled echo T1 and fast spin-echo T2-weighted images pre and post MnDPDP (0.005 mmol/kg b.w.) i.v. administration. Analysis could be performed in 141 of 151 of the patients. The variables used in both single variable analysis and DA included contrast-to-noise ratios pre and post MnDPDP, presence of rim enhancement, margin, and heterogeneity of a lesion pre and post MnDPDP. The classification of diagnoses using DA was compared with a standard of reference (HCC in 23%, metastases in 25%, cyst in 13%, FNH in 10%, hemangioma in 11%, and other or no lesion in 18% of the patients; histology in 49%, long-term follow-up in 51% of the cases). In the differentiation of the various hepatic lesions, CNR together with the presence of heterogeneity or rim enhancement as variables for DA gave the highest sensitivity, specificity, and accuracy which ranged between 65 and 93, 44 and 83, and 65 and 86%, respectively. The DA models based on post-MnDPDP variables showed better classification results than the models based on pre-MnDPDP variables. An improvement of accuracy was observed when differentiating HCC from FNH lesion groups (48.9-67.4%; p < or = 0.05), and when differentiating HCC from metastasis lesion groups (68.3-84.1%; p < or = 0.01). In all regards there was no difference for T2-weighted images pre and post MnDPDP. By combining quantitative and qualitative variables, DA proved to be a useful tool in lesion discrimination. Due to considerable heterogeneity within some of the lesion type groups, the definite diagnostic impact of MnDPDP cannot be completely established yet, and further investigation is still necessary.
本研究的目的是确定使用计算机多变量判别分析（DA），在注射锰福地匹三钠（MnDPDP）前后，能否通过特定的定量和定性成像特征来区分不同类型的局灶性肝病变。在一项多中心试验中，对151例患有局灶性肝病的患者进行了研究，在静脉注射MnDPDP（0.005 mmol/kg体重）前后，分别于1.5T和1.0T磁场下采用梯度回波T1加权像和快速自旋回波T2加权像进行检查。151例患者中有141例可进行分析。单变量分析和判别分析中使用的变量包括MnDPDP注射前后的对比噪声比、边缘强化的存在情况、病变的边界以及MnDPDP注射前后病变的异质性。将判别分析得出的诊断分类结果与参考标准进行比较（23%为肝细胞癌，25%为转移瘤，13%为囊肿，10%为局灶性结节性增生，11%为血管瘤，18%为其他病变或无病变；49%经组织学检查，51%经长期随访）。在区分各种肝病变时，以对比噪声比以及异质性或边缘强化的存在情况作为判别分析的变量，其敏感性、特异性和准确性最高，分别在65%至93%、44%至83%以及65%至86%之间。基于MnDPDP注射后变量的判别分析模型比基于注射前变量的模型分类结果更好。在区分肝细胞癌与局灶性结节性增生病变组时（48.9% - 67.4%；p≤0.05）以及区分肝细胞癌与转移瘤病变组时（68.3% - 84.1%；p≤0.01），准确性有所提高。在各方面，MnDPDP注射前后的T2加权像均无差异。通过结合定量和定性变量，判别分析被证明是一种有用的病变鉴别工具。由于某些病变类型组内存在相当大的异质性，MnDPDP的确切诊断价值尚未完全确立，仍需进一步研究。
