基于3D区域增长的磁共振水脂分离方法

Three-dimensional region growing method-based water-fat separation in magnetic resonance image

目的:为提高磁共振水脂分离结果的准确性,提出一种基于3D区域增长的水脂分离方法,使得水脂分离结果更稳定和准确。方法:利用相位平滑性假设,引入图像中层与层之间的相似性信息,提高相位估计的准确性,并且在最后通过3D迭代平滑操作,获得最终的相位图,利用最小二乘法得到水脂分离结果。结果:采集4名志愿者腹部数据,在与基于2D区域增长方法的对比实验中,本方法能够准确分离出水和脂肪,而基于2D区域增长的方法发生了不同程度的水脂分离错误。应用本方法对4组数据共100层进行水脂分离实验,在100层的水脂分离结果中均无分离错误。结论:基于3D区域增长算法能明显改善水脂分离结果的准确性,可为临床诊断提供更准确的辅助信息。

Objective To achieve a stable and accurate water-fat separation by a method of water-fat separation based on threedimensional（3D） region growing method. Methods We assumed that the phase was smoothness in the phase map, and used the information of neighborhood pixels in different slices to improve the accuracy of the phase estimation. The final phase map was obtained by 3D iterative smoothing, and the water-fat separation was achieved with the least squares method. Results The abdomen data of 4 volunteers were collected. The proposed method obtained an accurate water-fat separation which was better than that obtained with two-dimensional region growing method. Some errors were found in the results obtained with twodimensional region growing method, however, the proposed method for 4 groups of data（100 slices） didn＇t produced any errors in final water-fat separation. Conclusion The 3D regional growth algorithm improves the accuracy of water-fat separation,providing more accurate information for the clinical diagnosis.