宫颈癌磁共振扩散加权成像表观扩散系数：熵值的开发及初步应用

Development and primary application of entropy based on apparent diffusion coefficient mapping of magnetic resonance diffusion weighted imaging for cervical cancer

目的:基于宫颈癌磁共振扩散加权成像(DWI)表观扩散系数(ADC)参数图计算熵值,并评估其初步临床应用。方法:前瞻性选取27例从2013年10月至2014年5月病理证实为宫颈癌患者,行盆腔3.0 T磁共振成像,由两名具有5年以上磁共振读片经验的放射科医师使用本课题组自主研发的医学影像分析软件(Image Analyzer 1.0),在DWI图像上逐层勾画宫颈癌肿瘤组织及周围正常宫颈组织,获取感兴趣容积,并采用两种不同方法计算ADC熵值(一阶熵及二阶熵)。使用配对样本t检验比较宫颈癌肿瘤组织和周围正常宫颈组织的熵值有无显著性差异,采用受试者操作特征(ROC)曲线分析熵值在宫颈癌中的诊断效能,并采用组内相关系数(ICC)进行观察者间一致性分析。结果:宫颈癌肿瘤组织的ADC一阶熵(5.74±0.62)和二阶熵(10.25±1.65)均显著高于周围正常宫颈组织的ADC一阶熵(4.90±0.48)和二阶熵(8.03±0.78)(P<0.000 1)。ADC一阶熵和二阶熵的ROC曲线下面积分别为0.868和0.875,一阶熵(ICC=0.981)与二阶熵(ICC=0.985)均具有很高的观察者间一致性。结论:基于磁共振DWI成像ADC参数图开发的熵值有助于宫颈癌的评估。

Objective To evaluate the primary clinical application of entropy based on apparent diffusion coefficient（ADC）mapping of magnetic resonance（MR） diffusion weighted imaging（DWI） for cervical cancer. Methods From October 2013 to May 2014, 27 patients with cervical cancers confirmed by pathology were prospectively selected and underwent pelvic MR-DWI scans on a 3.0 T scanner. The self-compiled image analysis software（Image Analyzer 1.0） was applied by two radiologists who had above 5 years’ experience in gynecological MR imaging to manually obtain volumes of interests（VOI） of cervical cancer tissues and surrounding normal cervical tissues segmented on each slice of DWI image. ADC entropies, including first-order entropy and second-order entropy, were calculated by two different methods. Paired-sample t test was used to compare the entropies between cervical cancer tissues and surrounding normal cervical tissues; receiver operating characteristic（ROC） curves were conducted to assess the efficiency of entropy in the diagnosis of cervical cancer; intraclass correlation coefficient（ICC）was used for inter-observer consistency analysis. Results The ADC first-order entropy（5.74±0.62） and second-order entropy（10.25±1.65） of cervical cancer tissues were significantly higher than those of surrounding normal cervical tissues, respectively（4.90±0.48） and（8.03±0.78）（P〈0.000 1）. ADC first-order entropy had an area under receiver operating characteristic curve（AUC）of 0.868, and ADC of second-order entropy had an AUC of 0.875. Both first-order entropy（ICC=0.981） and second-order entropy（ICC=0.985） had excellent inter-observer consistency. Conclusion Entropy based on ADC mapping of MR-DWI has considerable help for assessing cervical cancer.