基于多b值扩散加权成像的多模型拟合在胶质瘤术前分级中的价值

发布时间：2018-05-15 20:09

本文选题：胶质瘤 + 磁共振　；参考：《南方医科大学》2017年博士论文

【摘要】：目的DWI序列相关参数评价胶质瘤有很高价值,但是不同的拟合模型参数诊断之间有很大差异,不同的ROI勾画方法也有很大差异。我们的研究目的是探讨哪些参数、哪种ROI勾画方法在胶质瘤术前分级中更有价值。方法46例经病理证实的胶质瘤患者,术前行磁共振常规序列检查、波谱检查、多b值扩散加权成像、DSC灌注检查和常规序列增强。多b值DWI序列进行单指数、双指数、拉伸指数和三指数的多模型拟合,ROI勾画采用Hot spot法、2D法和3D法三种。比较各拟合模型参数不同ROI勾画方法在术前鉴别胶质瘤高级别组与低级别组中的价值,筛选出敏感性和特异性较高的方案。同时比较各参数在鉴别胶质瘤各级之间的价值,筛选最佳方案。结果高级别与低级别胶质瘤波谱参数比较(以read1为例),肿瘤侧代谢物参数在高级别与低级别组之间比较无统计学差异(P=0.412,NAA/Cr;P=0.148,Cho/Cr;P=0.306,Cho/NAA)。DSC灌注参数比较,肿瘤侧高级别组与低级别组比较,有明显统计学差异(P=0.042,rCBF-+CMax;P=0.030,rCBF-rCBFMax;P=0.002,rCBV-+CMax;P=0.001,rCBV-rCBVMax;P=0.001,MTT-+CMax;P=0.000,MTT-MTTMax)。单指数模型拟合Hot spot ROI比较,ADC-ADCmin值在高级别组与低级别组间有差异(P=0.001),ADC-+CMax值在组间比较无差异(P=0.273);2D ROI相关参数比较,ADC-P10-2D在组间有统计学差异(P=0.035),其余参数无统计学差异。3DROI 相关参数比较,ADC-mean-3D、ADC-P5-3D 和 ADC-P10-3D 在组间有统计学差异(P=0.041,ADC-mean-3D;P=0.009,ADC-P5-3D;P=0.005,ADC-P10-3D),其余参数无统计学差异。双指数模型拟合HotspotROI比较,D-Dmin值和f-fMax在高级别组与低级别组间有差异(P=0.000,D-Dmin;P=0.002,f-fMax),其余参数在组间比较无差异;2D ROI相关参数比较,D-P5-2D、D-P10-2D、f-mean-2D、f-median-2D、f-P90-2D 和 f-skewness-2D 在组间有统计学差异(P=0.039,D-P5-2D;P=0.022,D-P10-2D;P=0.001,f-mean-2D;P=0.000,f-median-2D;P=0.016,f-P90-2D;P=0.014,f-skewness-2D),其余参数无统计学差异;3D ROI相关参数比较,D-mean-3D、D-P5-3D、D-P10-3D、f-mean-3D、f-median-3D 和 f-P90-3D 在组间有统计学差异(P=0.025,D-mean-3D;P=0.002,D-P5-3D;P=0.002,D-P10-3D;P=0.000,f-mean-3D;P=0.000,f-median-3D;P=0.043,f-P90-3D),其余参数无统计学差异。拉伸指数模型拟合Hot spot ROI比较,DDC-DDCmin值和α-αmin在高级别组与低级别组间有差异(P=0.001,DDC-DDCmin;P=0.000,α-αmin),其余参数在组间比较无差异;2D ROI相关参数比较,DDC-P5-2D、DDC-P10-2D、α-mean-2D 和 α-median-2D 在组间有统计学差异(P=0.033,DDC-P5-2D;P=0.017,DDC-P10-2D;P=0.009,α-mean-2D;P=0.004,α-median-2D),其余参数无统计学差异;3D ROI相关参数比较,DDC-mean-3D、DDC-median-3D、DDC-P5-3D、DDC-P10-3D、α-mean-3D、α-median-3D、α-P5-3D 和 α-P10-3D 在组间有统计学差异(P=0.067,DDC-mean-3D;P=0.047,DDC-median-3D;P=0.006,DDC-P5-3D;P=0.005,DDC-P10-3D;P=0.005,α-mean-3D;P=0.006,α-median-3D;P=0.033,α-P5-3D;P=0.014,α-P10-3D),其余参数无统计学差异。三指数模型拟合Hot spot ROI比较,Ff-+CMax值、Ds-Dsmin和Ff-FfMax在高级别组与低级别组间有差异(P=0.010,Ff-+CMax;P=0.005,Ds-Dsmin;P=0.007,Ff-FfMax),其余参数在组间比较无差异;2DROI相关参数比较,Ds-P5-2D、Ds-P10-2D、Ff-mean-2D、Ff-median-2D 和Ff-skewness-2D 在组间有统计学差异(P=0.030,Ds-P5-2D;P=0.009,Ds-P10-2D;P=0.007,Ff-mean-2D;P=0.015,Ff-median-2D;P=0.003,Ff-skewness-2D),其余参数无统计学差异;3D ROI相关参数比较,Ds-P5-3D、Ds-P10-3D、Ff-mean-3D、Ff-median-3D、Ff-skewness-3D、Fp-P90-3D、和 Fp-P95-3D 在组间有统计学差异(P=0.017,Ds-P5-3D;P=0.014,Ds-P10-3D;P=0.036,Ff-mean-3D;P=0.003,Ff-median-3D;P=0.013,Ff-skewness-3D;P=0.016,Fp-P90-3D;P=0.017,Fp-P95-3D),其余参数无统计学差异。用对侧白质区内部校正后参数与校正前参数诊断效能比较,Hot spot ROI法比较,校正后参数诊断效能高于校正前;2D法和3D法校正后参数诊断效能低于校正前。各相关参数信度分析,2D和3D法相关参数信度高于Hot spot法,3D法略高于2D法;DSC灌注相关参数中以rCBV值组内相关系数最高(ICC=0.994,rCBV-+CMax;ICC=0.994,rCBV-rCBVMax);单指数模型拟合参数中以ADC-P10-3D和ADC-P5-3D组内相关系数最高(ICC=0.994,ADC-P10-3D;ICC=0.982,ADC-P5-3D);双指数模型拟合参数中以 f-mean-3D 和 f-median-2D 组内相关系数最高(ICC=0.999,f-median-3D;ICC=0.998,f-median-2D);拉伸指数模型拟合参数中以α-median-3D和α-median-2D 组内相关系数最高(ICC=0.997,α-median-3D;ICC=0.997,α-median-2D);三指数模型拟合参数中以Ff-median-3D和Ff-median-2D组内相关系数最高(ICC=0.997,Ff-median-3D;ICC=0.997,Ff-median-2D)。各参数与胶质瘤病理分级相关性分析,DSC灌注参数中以rCBV-rCBVMax相关性最高(r=0.695);单指数模型中以ADC-P5-3D相关性最高(r=-0.456);双指数模型中以f-fMax相关性最高(r=0.507);拉伸指数模型中以α-αmin相关性最高(r=-0.607);三指数模型中以Ff-+CMax相关性最高(r=0.437)。多模型拟合参数与DSC灌注参数相关性分析,以f-median-2D与rCBV-rCBVMax相关性最高(r=0.604);多模型拟合参数中扩散相关参数以ADC Vs DDC(r=0.968)和D Vs Ds(r=0.966)相关性最强。各参数鉴别不同级别能力比较,波谱参数在各级之间无统计学差异;DSC灌注相关参数中,rCBV-+CMax、rCBV-rCBVMax、MTT-+CMax 和 MTT-MTTMax 在各级之间有统计学差异(P=0.006,rCBV-+CMax;P=0.002,rCBV-rCBVMax;P=0.001,MTT-+CMax;P=0.000,MTT-MTTMax);单指数模型拟合相关参数中,ADC-ADCmin、ADC-P10-2D、ADC-P5-3D和ADC-P10-3D 在各级之间有统计学差异(P=0.003,ADC-ADCmin;P=0.046,ADC-P10-2D;P=0.009,ADC-P5-3D;P=0.015,ADC-P10-3D);双指数模型拟合相关参数中,D-Dmin、f-fMax、D-P10-2D、f-mean-2D、f-median-2D、f-skewness-2D、D-P10-3D、f-mean-3D 和 f-median-3D 在各级之间有统计学差异(P=0.002,D-Dmin;P=0.003,f-fMax;P=0.049,D-P10-2D;P=0.002,f-mean-2D;P=0.001,f-median-2D;P=0.032,f-skewness-2D;P=0.002,D-P5-3D;P=0.005,D-P10-3D;P=0.002,f-mean-3D;P=0.000,f-median-3D);拉伸指数模型拟合相关参数中,DDC-DDCmin、α-αmin、DDC-P10-2D、α-mean-2D、α-median-2D、DDC-P5-3D、DDC-P10-3D、α-mean-3D 和 α-median-3D 在各级之间有统计学差异(P=0.005,DDC-DDCmin;P=0.000,α-αmin;P=0.045,DDC-P10-2D;P=0.021,α-mean-2D;P=0.012,α-median-2D;P=0.006,DDC-P5-3D;P=0.013,DDC-P10-3D;P=0.016,α-mean-3D;P=0.025,α-median-3D);三指数模型拟合相关参数中,Ff-+CMax、Ds-Dsmin、Ff-FfMax、Ds-P10-2D、Ff-mean-2D、Ff-median-2D、Ff-skewness-2D、Ds-P5-3D、Ds-P10-3D 和 Ff-median-3D 在各级之间有统计学差异(P=0.014,Ff-+CMax;P=0.018,Ds-Dsmin;P=0.021,Ff-FfMax;P=0.020,Ds-P10-2D;P=0.026,Ff-mean-2D;P=0.013,Ff-median-2D;P=0.016,Ff-skewness-2D;P=0.029,Ds-P5-3D;P=0.045,Ds-P10-3D;P=0.011,Ff-median-3D);各有价值参数在各级之间两两比较,Ⅱ级与Ⅲ级之间有统计学差异;Ⅱ级与Ⅳ级之间有统计学差异,但是Ⅲ级与Ⅳ级之间无统计学差异。结论1、多b值DWI序列中不同的拟合方案中有价值的参数不同,要根据选择的拟合方案选择合适的参数,以双指数模型诊断价值最高;2、各拟合方案中不同的ROI勾画方法有价值参数不同,要根据不同的ROI勾画方法选择不同的参数,以2D方法的诊断价值最高;3、各参数信度分析3D法相关参数最高,2D法次之,Hot spot法最低;4、各种相关参数中f值与DSC灌注相关参数相关性最高,能反映肿瘤内的灌注信息;4、弥散相关参数中,ADC值与DDC值相关性最高,D值与Ds值相关性最高;6、强化最明显区域各参数值的诊断价值不及各参数图极值区域参数值;7、各有价值参数能很好的鉴别Ⅱ级与Ⅲ级、Ⅱ级与Ⅳ级胶质瘤,但是鉴别Ⅲ级与Ⅳ级较难;8、三指数拟合模型相关参数在胶质瘤术前分级中有很大价值。
[Abstract]:Objective DWI sequence related parameters are of high value in evaluating glioma, but there is a great difference between the parameters of different fitting models and different ROI drawing methods. The purpose of this study is to discuss which parameters and which ROI method is more valuable in the preoperative grading of glioma. Method 46 cases of the pathologically confirmed glue Patients with stromal tumors underwent conventional magnetic resonance imaging, spectrum examination, multi B diffusion weighted imaging, DSC perfusion examination and routine sequence enhancement. Multiple b value DWI sequences were fitted with multiple models of single index, double index, tensile index and three index, and Hot spot method, 2D method and 3D method were used in ROI delineation. The parameters of each model were compared with ROI hook. The value of the high grade and low grade group of glioma was identified before the operation, and the high sensitivity and specificity were screened. At the same time, the value of each parameter in the identification of glioma was compared and the best scheme was screened. Results the ratio of high and low grade glioma spectrum parameters was compared with read1, and the parameter of tumor side metabolite was in the case of the high grade and low grade glioma. There was no statistical difference between the high and low level groups (P=0.412, NAA/Cr; P=0.148, Cho/Cr; P=0.306, Cho/NAA).DSC perfusion parameters, and there was a significant difference between the high level group and the low level group (P=0.042, rCBF-+CMax; P=0.030, rCBF-rCBFMax; P=0.002) TT-MTTMax). Compared with Hot spot ROI, the ADC-ADCmin value is different between the advanced group and the low level group (P=0.001), and the ADC-+CMax value is no difference between the groups (P=0.273); the 2D ROI related parameters are compared, and the ADC-P10-2D is statistically different between the groups (P=0.035), and the other parameters are not statistically different. 3D, ADC-P5-3D and ADC-P10-3D were statistically different between groups (P=0.041, ADC-mean-3D; P=0.009, ADC-P5-3D; P=0.005, ADC-P10-3D), and the rest of the parameters were not statistically different. The double exponential model fitted HotspotROI, D-Dmin and f-fMax were different between the advanced and low-level groups, and the rest of the parameters were in the inter group ratio. 2D ROI related parameters, D-P5-2D, D-P10-2D, f-mean-2D, f-median-2D, f-P90-2D and f-skewness-2D were statistically different between the groups (P=0.039, D-P5-2D; P=0.022, D-P10-2D; dialectical; dialectical; excluded), the other parameters were not statistically different; D-mean-3D, D-P5-3D, D-P10-3D, f-mean-3D, f-median-3D and f-P90-3D have statistical differences between the groups (P=0.025, D-mean-3D; P=0.002, D-P5-3D; P=0.002, D-P10-3D), and the rest of the parameters are not statistically different. There was a difference between the advanced group and the low level group (P=0.001, DDC-DDCmin; P=0.000, alpha - alpha min), and the other parameters were not different between the groups. The 2D ROI related parameters, DDC-P5-2D, DDC-P10-2D, alpha -mean-2D and alpha -median-2D were statistically different between the groups. There is no statistical difference in the residual parameters; DDC-mean-3D, DDC-median-3D, DDC-P5-3D, DDC-P10-3D, alpha -mean-3D, alpha -median-3D, alpha -P5-3D and alpha -P10-3D are statistically different in the 3D ROI correlation parameters. 33, alpha -P5-3D; P=0.014, alpha -P10-3D), the other parameters are not statistically different. The three index model fits the Hot spot ROI comparison, Ff-+CMax value, Ds-Dsmin and Ff-FfMax are different between the advanced group and the lower class group (P=0.010, Ff-+CMax; P=0.005, etc.), and the other parameters are not different between the groups. S-P10-2D, Ff-mean-2D, Ff-median-2D and Ff-skewness-2D have statistical differences between groups (P=0.030, Ds-P5-2D; P=0.009, Ds-P10-2D; P=0.007, Ff-mean-2D; P=0.015, Ff-median-2D; Ff-skewness-2D). 3D, and Fp-P95-3D were statistically different between groups (P=0.017, Ds-P5-3D; P=0.014, Ds-P10-3D; P=0.036, Ff-mean-3D; P=0.003, Ff-median-3D; P=0.013, Ff-skewness-3D; Ff-skewness-3D. The diagnostic efficiency of the corrected parameters is higher than that before the correction. The parameter diagnostic efficiency of the 2D and 3D methods is lower than that before the correction. The reliability of the relevant parameters is higher than the Hot spot method, and the 3D method is slightly higher than the 2D method, and the correlation coefficients of the DSC perfusion parameters are the highest in the rCBV value group. CBVMax); the correlation coefficients of the ADC-P10-3D and ADC-P5-3D groups in the single exponential model were the highest (ICC=0.994, ADC-P10-3D; ICC=0.982, ADC-P5-3D), and the correlation coefficients of the f-mean-3D and f-median-2D groups in the double exponential model were the highest (ICC=0.999, f-median-3D; ICC=0.998 and ADC-P5-3D), and the parameters of the tensile index model were alpha - The correlation coefficient of median-3D and alpha -median-2D was the highest (ICC=0.997, alpha -median-3D; ICC=0.997, alpha -median-2D). The correlation coefficient of Ff-median-3D and Ff-median-2D in the three exponential model was the highest (ICC=0.997, Ff-median-3D; ICC=0.997, Ff-median-2D). The correlation between the parameters and the pathological grading of glioma was analyzed. The rCBV-rCBVMax correlation is the highest (r=0.695); in the single exponential model, the ADC-P5-3D correlation is the highest (r=-0.456); the f-fMax correlation is the highest in the double exponential model (r=0.507); the tensile index model is the highest (r=-0.607) in the alpha - a min correlation; the three index model is the highest (r=0.437) in the Ff-+ CMax correlation. The correlation between f-median-2D and rCBV-rCBVMax is the highest (r=0.604), and the correlation parameter of the multiple model fitting parameters is the strongest correlation between the ADC Vs DDC (r=0.968) and D Vs Ds (r=0.966). There are statistical differences between X, MTT-+CMax and MTT-MTTMax at all levels (P=0.006, rCBV-+CMax; P=0.002, rCBV-rCBVMax; P=0.001, MTT-+CMax; P=0.000, MTT-MTTMax). 9, ADC-P5-3D; P=0.015, ADC-P10-3D); the two exponential models fit the relevant parameters, D-Dmin, f-fMax, D-P10-2D, f-mean-2D, f-median-2D, f-skewness-2D, D-P10-3D, f-mean-3D and f-median-3D. S-2D; P=0.002, D-P5-3D; P=0.005, D-P10-3D; P=0.002, f-mean-3D; P=0.000, f-median-3D); DDC-DDCmin, alpha min in the tensile exponential model. DDC-P10-2D; P=0.021, alpha -mean-2D; P=0.012, alpha -median-2D; P=0.006, DDC-P5-3D; P=0.013, DDC-P10-3D; P=0.016, alpha -mean-3D; P=0.025, and alpha. P=0.014 (Ff-+CMax; P=0.018, Ds-Dsmin; P=0.021, Ff-FfMax; P=0.020, Ds-P10-2D; P=0.026, Ff-mean-2D; P=0.013, Ff-median-2D; Ff-median-2D); there are statistical differences between the value parameters at all levels, between class II and grade III; class II and There are statistical differences between grade IV, but there is no statistical difference between grade III and grade IV. Conclusion 1, the value of different parameters in different fitting schemes of multiple b value DWI sequences are different. It is necessary to select the appropriate parameters according to the selected fitting scheme, and the value of the double exponential model is the highest. 2, the different ROI drawing methods in each fitting scheme have value reference. Different parameters should be selected according to different ROI methods, and the diagnostic value of 2D method is the highest. 3, the parameters reliability analysis 3D method has the highest correlation parameters, 2D method is the second, Hot spot method is the lowest; 4, the correlation parameter of F and DSC perfusion parameters is most high, can reflect the perfusion information within the tumor; 4, dispersion related parameters The correlation between the ADC value and the DDC value is the highest, the D value has the highest correlation with the Ds value; 6, the diagnostic value of each parameter in the most obvious region is less than the parameter value of the extreme value of the parameter graphs; 7, each valuable parameter can identify the grade II and III, grade II and grade IV glioma, but the discrimination of grade III and IV is more difficult; the 8, three index fitting model parameters It is of great value in preoperative grading of glioma.

【学位授予单位】：南方医科大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：R445.2;R739.41

【相似文献】