一种快速的全自动超声子宫图像分割算法

使用超声成像进行子宫节育环检查工作已在我国广泛地开展,利用图像识别技术进行计算机辅助诊断对于减轻检查人员工作负担意义十分明显,其中图像分割部分的主要目标则是快速地全自动分割开图中的几个主要器官及节育环。本研究提出了一种快速的全自动子宫图像分割算法。该算法包括以下三个主要步骤:首先运用BP神经网络处理图像整体灰度分布获取基准分割阈值;其后使用超声图像斑点噪声统计特征进行同质区域判别,并根据局部灰度分布自适应调整分割阈值;最后使用数学形态学算子对分割效果做进一步的改善。基于由1200幅超声子宫图像组成的图像库,对所提算法与最大类别方差法、SNAKE活动轮廓模型等数种常用分割算法进行了性能比较,实验结果表明所提算法在速度与准确程度两方面均表现良好,平均耗时为0.93s/幅,准确程度达到了94%。本算法无需人工干预,分割速度快,分割准确程度能够被临床医生所接受,可以用作超声子宫图像辅助诊断系统的图像分割部分,具有很好的应用前景。

A Fast Automatic Segmentation Algorithm on the Ultrasound Uterus Images

The intrauterus device examinations by ultrasound imaging are widely used in China.For alleviating the working burden resulting from manual recognition,the assistant diagnosis system,based on automatic image recognition,is of great importance.In this paper,a fast automatic segmentation algorithm was proposed.Firstly,a baseline threshold was obtained by the BP neural network using the global brightness distribution;Secondly,the homogeneity of the local region was analyzed and the threshold was adjusted under the local brightness distribution;Thirdly,the segmentation result was modified by the method of mathematical morphology.Based on the image library composed of 1200 ultrasound uterus images,we compared the proposed algorithm with several popular segmentation algorithms such as OTSU and SNAKE models.The results showed that the proposed algorithm performed very well both in speed and accuracy with the average time-consuming of 0.93 second per frame and the accuracy of 94%.The proposed algorithm is promising to be the segmentation part of the assistant diagnosis system of ultrasound uterus images.