Hui Steve C N, Zhang Teng, Shi Lin, Wang Defeng, Ip Chei-Bing, Chu Winnie C W
Department of Imaging and Interventional Radiology, Prince of Wales Hospital, The Chinese University of Hong Kong, Sha Tin, New Territories, Hong Kong.
Department of Medicine and Therapeutics, Prince of Wales Hospital, The Chinese University of Hong Kong, Sha Tin, New Territories, Hong Kong; Chow Yuk Ho Technology Centre for Innovative Medicine, Prince of Wales Hospital, The Chinese University of Hong Kong, Sha Tin, New Territories, Hong Kong.
Magn Reson Imaging. 2018 Jan;45:97-104. doi: 10.1016/j.mri.2017.09.016. Epub 2017 Oct 7.
To develop a reliable and reproducible automatic technique to segment and measure SAT and VAT based on MRI.
Chemical-shift water-fat MRI were taken on twelve obese adolescents (mean age: 16.1±0.6, BMI: 31.3±2.3) recruited under the health monitoring program. The segmentation applied a spoke template created using Midpoint Circle algorithm followed by Bresenham's Line algorithm to detect narrow connecting regions between subcutaneous and visceral adipose tissues. Upon satisfaction of given constrains, a cut was performed to separate SAT and VAT. Bone marrow was consisted in pelvis and femur. By using the intensity difference in T2*, a mask was created to extract bone marrow adipose tissue (MAT) from VAT. Validation was performed using a semi-automatic method. Pearson coefficient, Bland-Altman plot and intra-class coefficient (ICC) were applied to measure accuracy and reproducibility.
Pearson coefficient indicated that results from the proposed method achieved high correlation with the semi-automatic method. Bland-Altman plot and ICC showed good agreement between the two methods. Lowest ICC was obtained in VAT segmentation at lower regions of the abdomen while the rests were all above 0.80. ICC (0.98-0.99) also indicated the proposed method performed good reproducibility.
No user interaction was required during execution of the algorithm and the segmented images and volume results were given as output. This technique utilized the feature in the regions connecting subcutaneous and visceral fat and T2* intensity difference in bone marrow to achieve volumetric measurement of various types of adipose tissue in abdominal site.
开发一种基于磁共振成像（MRI）的可靠且可重复的自动技术，用于分割和测量皮下脂肪组织（SAT）和内脏脂肪组织（VAT）。
对在健康监测项目中招募的12名肥胖青少年（平均年龄：16.1±0.6，体重指数：31.3±2.3）进行化学位移水脂MRI检查。分割过程应用了使用中点圆算法创建的辐条模板，随后采用布雷森汉姆线算法来检测皮下和内脏脂肪组织之间的狭窄连接区域。在满足给定约束条件后，进行切割以分离SAT和VAT。骨盆和股骨中包含骨髓。利用T2*中的强度差异，创建一个掩码以从VAT中提取骨髓脂肪组织（MAT）。使用半自动方法进行验证。应用皮尔逊系数、布兰德 - 奥特曼图和组内相关系数（ICC）来测量准确性和可重复性。
皮尔逊系数表明，所提出方法的结果与半自动方法具有高度相关性。布兰德 - 奥特曼图和ICC显示两种方法之间具有良好的一致性。在腹部较低区域的VAT分割中获得了最低的ICC，而其他区域均高于0.80。ICC（0.98 - 0.99）也表明所提出的方法具有良好的可重复性。
在算法执行过程中无需用户交互，并且输出分割图像和体积结果。该技术利用皮下和内脏脂肪连接区域的特征以及骨髓中的T2*强度差异，实现腹部不同类型脂肪组织的体积测量。
