弥散加权成像-表观扩散系数图纹理分析对腹膜后去分化及高分化脂肪肉瘤的鉴别诊断价值

Differential diagnostic value of DWI-ADC image texture analysis in retroperitoneal dedifferentiation and well-differentiated liposarcoma

目的:探讨弥散加权成像-表观扩散系数(DWI-ADC)图纹理分析法用于腹膜后去分化及高分化脂肪肉瘤的鉴别诊断价值。方法:回顾性分析2009年11月至2018年12月在复旦大学附属中山医院经手术病理确诊的腹膜后脂肪肉瘤患者。患者均行腹腔或盆腔MRI平扫、DWI及多期增强扫描检查。采用德国Siemens Multiparametric软件进行纹理分析,获得DWI-ADC图灰度直方图参数,包括Skewness、E.Kurtosis、DiffEntropy、Contrst、Entropy、ADC均值、ADC中位值、5%ADC值(P 5)、95%ADC值(P 95)、ADC三等分值(低、中、高)等。所有参数采用Shapiro-Wilk检验,符合正态分布的数据采用独立样本t检验比较,非正态分布的数据采用非参数Mann-Whitney U检验,对两组患者间有差异的变量行ROC曲线分析评估其鉴别诊断效能。结果:共纳入患者117例,其中去分化脂肪肉瘤患者36例,高分化脂肪肉瘤81例。两组患者间Volume、ADC均值、ADC中位值、P 95、Skewness、E.Kurtosis值差异有统计学意义(P<0.05)。E.Kurtosis、Skewness、ADC均值、ADC中位值、P 95、Volume的ROC曲线下面积分别为0.665、0.738、0.635、0.633、0.652、0.819,敏感度分别为63.89%、61.11%、83.33%、82.86%、62.27%、97.22%,特异度分别为77.78%、86.42%、53.75%、51.22%、66.71%、59.26%。结论:DWI-ADC图纹理分析法有助于腹膜后去分化与高分化脂肪肉瘤的鉴别诊断,其中Skewness、E.Kurtosis、Volume诊断效能较高。

Objective: Differential diagnosis of retroperitoneal dedifferentiated and well-differentiated liposarcoma using DWI-ADC image texture analysis. Methods: The patients with retroperitoneal liposarcoma confirmed by operation and pathology in Zhongshan Hospital, Fudan University from November 2009 to December 2018 were retrospectively analyzed, among which 36 cases were dedifferentiated liposarcoma, and 81 cases were highly differentiated liposarcoma, all patients were examined by abdominal or pelvic MR plain scan, DWI and multi-phase enhanced scan. Texture analysis was carried out by using Siemens multiparametric software of Germany. Gray histogram parameters such as Skewness, E. Kurtosis, DIffEntropy, Contrst Entropy, ADC mean, ADC median, 5% ADC value, 95% ADC value, and ADC tri-value(low, medium, and high) were obtained. Shapiro-Wilk test was used to test the normal distribution of the data, which accorded with the normal distribution. Independent sample t-test was used to compare the data of the non-normal distribution. Mann-Whitney U test was used to test the data of the non-normal distribution. The ROC curve of the variables that were different between the two groups was used to evaluate the differential diagnostic efficacy. Results: There were 117 patients. There was significant difference in the Volume, Mean, Median, P95, Skewness, and E.Kurtosis between the two groups(P<0.05). The area under ROC curve of E.Kurtosis, Skewness, Mean, Median, P95, and Volume was about 0.665, 0.738, 0.635, 0.633, 0.652, and 0.819, respectively. The sensitivity was about 63.89%, 61.11%, 83.33%, 82.86%, 62.27%, and 97.22%, and the specificity was about 77.78%, 86.42%, 53.75%, 51.22%, 66.71%, and 59.26%. Conclusions: DWI-ADC image texture analysis is helpful in the differential diagnosis of retroperitoneal dedifferentiation and well-differentiated liposarcoma, among which Skewness, Kurtosis, and Volume are more effective.