摘要
目的探讨DWI的单指数、双指数、拉伸指数模型成像诊断前列腺癌的价值。方法回顾性分析前列腺MRI检查序列完整且经超声引导下穿刺活检病理诊断的34例患者,其中前列腺癌18例,前列腺增生16例,经穿刺活检共获得的癌灶28个、基质增生灶36个、正常外周带区域30个。患者均行前列腺FSET2WI、FSET1WI和多b值DWI扫描。对应穿刺活检部位,在图像上绘制前列腺癌灶、基质增生灶和正常外周带区域ROI,测量单指数模型参数(ADC值)、双指数模型参数[包括慢速扩散系数(D)、快速扩散系数(D*)和快速扩散所占容积分数(f)]和拉伸指数模型参数f包括分布扩散系数(DDC)、扩散异质性指数(α)1,并记录不同b值的信号强度,绘制不同模型的各组织信号拟合曲线。采用调整r2值比较不同模型的曲线拟合效果,并采用方差分析比较不同模型调整r2值的差异。采用单因素方差分析比较不同组织间各模型参数的差异。采用ROC曲线评价3种模型参数鉴别前列腺良、恶性病变的效能。结果前列腺癌灶、基质增生灶和正常外周带区域的3种不同扩散模型间调整r2值差异均有统计学意义(P均〈0.05),其中双指数扩散加权模型、拉伸指数扩散加权模型的调整r2值均高于单指数模型(P均〈0.05)。前列腺癌灶、基质增生灶和正常外周带区域的ADC值分别为(0.852±0.169)×10^-3、(1.443±0.201)×10^-3、(1.994±0.184)×10^-3mm2/s,D值分男0为(0.658±0.151)×10^-3、(1.149±0.171)×10^-3、(1.689±0.238)×10^-3mm2/s,DDC值分另0为(0.618±0.258)×10^-3、(1.431±0.329)×10^-3、(2.134±0.213)×10^-3mm2/s,α值分别为0.725±0.075、0.773±0.056、0.847±0.075,差异均有统计学意义(P均〈0.05)。ADC、D、DDC、α值诊断前列腺癌的ROC曲线下面积分别为0.995、0.991、0.984、0.773,分别以1.100×10^-3mm2/s、0.900×10^-3mm2/s、1.100×10^-3mm2/s和0.727为界值,诊断的敏感度和特异度分别为92.86%、98.48%,92.86%、98.48%,100.00%、92.42%和57.14%、86.36%。结论拉伸指数模型和双指数模型的曲线拟合优于单指数模型,单指数模型参数ADC值诊断前列腺癌效能较高,拉伸指数模型和双指数模型可以作为单指数模型的补充。
Objective To evaluate the diagnostic value of mono-exponential, bi-exponential, and stretched-exponential models of DWI in prostate cancer. Methods A retrospective study were performed in 34 patients with pathologically confirmed prostate cancer (n=18) and benign prostatic hypertrophy (n=16). Twenty eight prostate cancer ROIs, 36 stromal benign prostatic hyperplasia ROIs and 30 normal peripheral zone ROIs were analyzed. FSE T2WI, FSE T1WI and muti-b DWI were performed in all patients. ROIs wereplaced within the proven prostate cancer, stromal benign prostatic hyperplasia and normal peripheral zone. The parameters of mono-exponential (ADC), bi-exponential [slow diffusion coefficient (D), fast diffusion coefficient (D*) and perfusion fraction (f)] and stretched-exponential models [distributed diffusion coefficient (DDC) and stretching parameter (α] were recorded. The mean signal intensities of every ROI at different b values were recorded in order to assess the goodness of fit for different models. The adjusted r2 was calculated to assess the goodness of fit for different models. Variance analysis and q test were used to compare the values of r2 among groups. One-way analysis of variance was used to compare the different parameters. ROC curve was performed to evaluate the diagnosis value of different parameters in prostate cancer. Results The adjusted r2 were all statistically different among three models in prostate cancer, stromal benign prostatic hyperplasia, and normal peripheral zone (all P〈 0.05). The r2 ofbi-exponential and stretched-exponential models achieved significantly better fitting of DWI signal than the mono-exponential model (all P〈0.05).The ADC value of prostate cancer, stromal benign prostatic hyperplasia and normal peripheral zone were ( 0.852±0.169 ) × 10^-3, ( 1.443 ±0.201 ) ×10^-3, ( 1.994 ±0.184 ) × 10^-3mm2/s, respectively. The D value were (0.658±0.151)×10^-3, ( 1.149±0.171 )×10^-3, ( 1.689±0.238)×10^-3mm2/s, respectively. DDC were (0.618±0.258) × 10^-3, ( 1.431 ±0.329) × 10^-3, (2.134±0.213)× 10^-3mm2/s, respectively, α were 0.725±0.075, 0.773±0.056, 0.847±0.075, respectively. All had statistical difference (all P〈 0.05). The area under curve (AUC) of ADC, D, DDC and α in prostate cancer diagnosis were 0.995, 0.991, 0.984, 0.773. When the cutoff were 1.100 × 10^-3mm2/s, 0.900 × 10^-3mm2/s, 1.100 ×10^-3mm2/s, and 0.727 for ADC, D, DDC, and α, the sensitivity and specificity were 92.86%, 98.48%; 92.86%, 98.48%; 100.00%, 92.42%; 57.14%, 86.36%; respectively. Conclusions Though stretched-exponential and bi-exponential models achieved better fitting of DWI signal than the mono-exponential model, the ADC shows equal diagnosis value compared with D and DDC. The bi-exponential and stretched-exponential model can serve a supplement to mono-exponential model.
出处
《中华放射学杂志》
CAS
CSCD
北大核心
2015年第11期838-842,共5页
Chinese Journal of Radiology
基金
国家自然科学基金(81171307)
关键词
前列腺肿瘤
前列腺增生
磁共振成像
Prostate neoplasms
Prostate hyperplasia
Magnetic resonance imaging