
浏览全部资源
扫码关注微信
1. 扬州大学临床医学院,江苏,扬州,225001
2. 苏北人民医院影像科,江苏,扬州,225001
网络出版:2019-12-28,
纸质出版:2019-12-28
移动端阅览
韩雷,石慧娴,吴刘洋,等. 基于扩散加权成像的纹理分析在胶质瘤分级中的价值[J]. 肿瘤影像学, 2019, 28(6): 358-364 https://doi.
org/10.19732/j.cnki.2096-6210.2019.06.002
韩雷,石慧娴,吴刘洋,等. 基于扩散加权成像的纹理分析在胶质瘤分级中的价值[J]. 肿瘤影像学, 2019, 28(6): 358-364 https://doi. DOI: 10.19732/j.cnki.2096-6210.2019.06.002.
org/10.19732/j.cnki.2096-6210.2019.06.002 DOI:
目的:
探讨基于扩散加权成像(diffusion-weighted imaging,DWI)的纹理分析在胶质瘤分级中的价值。
方法:
回顾性分析经病理学检查证实的30例低级别胶质瘤(low grade glioma,LGG)(Ⅱ级)与97例高级别胶质瘤(high grade glioma,HGG)(Ⅲ级46例,Ⅳ级51例),使用MaZda ver.4.6提取所有患者DWI图像中肿瘤实质信号最高区域的纹理特征并分析其中的直方图参数,包括均值、方差、偏度和峰度,以及第1、10、50、90和99百分位数(Pere.1%、Pere.10%、Pere.50%、Pere.90%、Pere.99%),方差、峰度这两个参数的比较采用非参数秩和检验(Mann-Whitney
U
检验),其余参数采用独立样本t检验(independent-samples
t
test),使用受试者工作特征(receiver operating characteristic,ROC)曲线分析有统计学意义的参数在HGG与LGG之间的诊断效能,运用多变量Logistic回归分析对有统计学意义的纹理参数进行建模并绘制ROC曲线评价模型效能。
结果:
直方图参数中的均值、方差与第1、10、50、90、99百分位数在LGG和HGG间的差异有统计学意义[(144.19847.133)
vs
(185.60940.341),(28.10139.529)
vs
(160.143211.832),(134.23343.673)
vs
(162.57740.478),(138.10044.970)
vs
(172.81439.384),(144.40047.211)
vs
(186.24740.473),(149.83349.537)
vs
(197.44342.977),(152.33350.384)
vs
(201.36143.720),
P
均<0.001],偏度、峰度在两组间的差异无统计学意义[(-0.3220.499)
vs
(-0.3690.542),
P
=0.669;(-0.1710.587)
vs
(-0.1350.973),
P
=0.440]。两组间的方差以29.23为阈值时具有最高的诊断效能,对应的灵敏度、特异度及曲线下面积(area under the curve,AUC)分别为72.16%、76.67%、0.793。通过这7个纹理参数建立的多
参数Logistic回归诊断模型的灵敏度、特异度及AUC分别为61.86%、86.67%、0.807。
结论:
基于DWI的纹理分析中的直方图参数可于术前有效地鉴别HGG与LGG,其中方差具有较高的诊断效能。
Objective:
To investigate the value of texture analysis based on diffusion-weighted imaging (DWI) in grading glioma.
Methods:
A retrospective analysis was made of 30 low grade glioma (LGG) (grade Ⅱ) and 97 high grade glioma (HGG) (grade Ⅲ 46 cases
grade Ⅳ 51 cases) confirmed by pathology. MaZda ver.4.6 was used to extract texture features of the highest signal region of tumor parenchymal in all patients DWI images and analyze histogram parameters
including mean
variance
skewness
kurtosis and Pere.1%
Pere.10%
Pere.50%
Pere.90%
Pere.99%. Mann-Whitney
U
was used to compare the variance and kurtosis
and independent-samples
t
test was used for the other parameters. Receiver operating characteristic (ROC) curve was used to analyze the diagnostic efficiency of statistically significant parameters between LGG and HGG. Multivariate Logistic regression analysis was used to model texture parameters with statistical significance and ROC curve was used to evaluate the efficiency of the model.
Results:
The mean
variance
Pere.1%
Pere.10%
Pere.50%
Pere.90%
Pere.99% of histogram parameters were significantly different between LGG and HGG [(144.19847.133)
vs
(185.60940.341)
(28.10139.529)
vs
(160.143211.832)
(134.23343.673)
vs
(162.57740.478)
(138.10044.970)
vs
(172.81439.384)
(144.40047.211)
vs
(186.24740.473)
(149.83349.537)
vs
(197.44342.977)
(152.33350.384)
vs
(201.36143.720)
all
P
<0.001]
while no significant difference in skewness or kurtosis between the two groups [(-0.3220.499)
vs
(-0.3690.542)
P
=0.669; (-0.1710.587)
vs
(-0.1350.973)
P
=0.440]. When the variance between the two groups was 29.23 as the threshold
the diagnostic ef
ficiency was the highest
and the corresponding sensitivity
specificity and area under the curve (AUC) were 72.16%
76.67% and 0.793
respectively. The sensitivity
specificity and AUC of the multiparameter Logistic regression diagnostic model established by these 7 texture parameters were 61.86%
86.67% and 0.807
respectively.
Conclusion:
The histogram parameters based on DWI texture analysis can effectively differentiate LGG from HGG before operation
among which variance has high diagnostic efficiency.
0
浏览量
1571
下载量
4
CNKI被引量
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024621