DWI单指数和双指数模型在前列腺癌Gleason分级中的应用:基于三维立体直方图分析

目的基于三维立体直方图分析,比较扩散加权成像(diffusion weighted imaging,DWI)单指数和双指数模型定量参数在前列腺癌Gleason分级中的应用价值。材料与方法回顾性纳入57例前列腺癌根治患者,术前均行MRI检查。DWI图像包含14个b值(0～1500 s/mm~2)。在DWI原始图像上围绕全肿瘤病灶逐层手动绘制感兴趣区(region of interest,ROI),并生成单指数和体素内不相干运动(intravoxel incoherent motion,IVIM)模型定量参数:表观扩散系数(apparent diffusion coefficient,ADC)、单纯水扩散系数(D)、灌注相关扩散系数(D*)、灌注分数f的直方图指标10th、25th、50th、75th、90th、mean、min、kurtosis、skewness,其与Gleason评分(Gleason score,GS)的相关性采用Spearman相关分析。将患者分成GS≤3+4与GS≥4+3组,两组间参数比较采用独立样本t检验。采用受试者工作特征(receiver operating characteristic curve,ROC)曲线分析评价各参数对低高级别前列腺癌的鉴别能力。P0.05为差异具有统计学意义。结果 GS≤3+4组23例,GS≥4+3组34例。ADC、D各直方图指标与GS呈负相关(ρ=-0.239～-0.411),其中90th ADC的相关系数最大。低级别组ADC、D值大于高级别组(P0.05)。90th ADC、90th D的曲线下面积(area under the roc curve,AUC)(分别为0.735和0.739)大于均值ADC和均值D,但差异无统计学意义(P值均0.05)。ADC、D的kurtosis与GS呈中等正相关(ρ分别为0.357和0.350,P0.05),skewness与GS无相关性。D*、f各直方图参数与GS无相关性,两组间差异无统计学意义。肿瘤体积与GS呈中等正相关(ρ=0.419,P=0.001),AUC(0.783)大于90th ADC、90th D,但各参数AUC两两相比,差异均无统计学意义(P均0.05)。结论 ADC、D各直方图参数对低级别和高级别前列腺癌有鉴别诊断价值,二者效能相当。D*、f不能预测前列腺癌的侵袭性。肿瘤体积有望成为前列腺癌病理分级的良好预测因子。

Comparison of apparent diffusion coefficient and intravoxel incoherent motion imaging in stratification of prostate cancer: based on three-dimensional whole-tumor histogram analysis

Objective: To investigate the three-dimensional whole-tumor histogram analysis of apparent diffusion coefficients (ADCs) and IVIM-DWI in discriminating the Gleason grades of prostate cancer. Materials and Methods: A total of 57 patients with prostate cancer (PCa) underwent MRI before radical prostatectomy. DWI was acquired at 3.0 T using a single-shot echo-planar imaging sequence with 14 b values (0—1500 s/mm2). A volume of interest was drawn manually around every tumor visible on the images. The 10th, 25th, 50th, 75th , 90th, mean, min, kurtosis, skewnessof ADC, D, D* and f were determined from a histogram that contained all voxels from all sections covering the entire tumor and correlated with the Gleason score (GS) by using the Spearman correlation coefficient (ρ). The patients were divided into GS≤3+4 and GS≥4+3 groups. The area under the receiver operating characteristic curve (AUC) was used to evaluate the ability of each parameter to differentiate the two groups. Results: GS≤3+4 in 23 cases (40.35%) and GS≥4+3 in 34 cases (59.65%). The 10th, 25th, 50th, 75th, 90th of ADC and D show a negative correlation with GS (ρ=-0.239—-0.411, P<0.05), and ρ values increased in turn. All above metrics were significantly different between two groups (P<0.05). The 90th ADC, 90th D show higher AUC (0.735 and 0.739, respectively) than mean ADC and mean D (0.719 and 0.698, respectively, P>0.05) in differentiating GS≤3+4 from GS≥4+3. There was no statistically significant difference between any two AUC values of ADC and D histogram metrics. The kurtosis of ADC and D were positively correlated with GS (ρ=0.357, P=0.006; ρ=0.350, P=0.008), but skewness had no correlation with GS. D* and f had no statistically significant correlation with GS either. The volume of whole-tumor lesion was positively correlated with GS (ρ=0.419, P=0.001), and showed a higher discriminatory performance (AUC=0.783) than ADC and D in differentiating GS≤3+4 from GS≥4+3. Conclusions: The histogram parameters of ADC and D had similar performance for prostate cancer stratification. D* and f were unable to distinguish between the different Gleason grades. Tumor volume may be a good predictor of PCa aggressiveness.