一种基于SSED的无规则噪声组织的去除方法

Segmentation Algorithm for Irregular Noise Tissue Based on Sequential Signed Euclidean Distance Transform

目的:在医学图像处理领域,往往会由于方案设计、算法选取等原因而产生大量的不规则的噪声组织。本文通过研究分析噪声组织形态特征,提出了一种新的快速去除噪声组织的方法,为医学图像处理的噪声组织去除领域提供一种全新的思路。方法:本文首次将距离映射应用到噪声组织去除领域,利用有符号串行欧氏距离映射(Signed Sequence Euclidean Distance Mapping,简称SSED)对图像进行距离映射计算,通过对距离映射图像进行适当的阈值化处理,实现了快速识别、去除各种不规则的噪声组织,与此同时确保了感兴趣目标区域的完整。结果:本方法已通过对12组临床心脏CT序列的实际测试,快速准确的去除了图像中存在的各种不规则的噪声组织,同时实现了对感兴趣目标区域的平滑。结论:该方法去除精度高、去除效率优于现有的形态学和骨架提取等各类方法,为临床医生的诊断提供了清晰且精确的辅助,大大提高了无规则噪声组织去除的效率和鲁棒性。

Objective In the field of medical image processing, there often remains much irregular noise tissue due to the projectdesign, the algorithm or other reasons. By studying morphological features of irregular noise tissue, we propose a new methodto quickly remove noise tissue and provide a tesh idea for noise removal field. Methods In this paper, we, for the first time,apply distance mapping to the field of noise tissue removal. The use of signed sequence Euclidean distance transform （SSED）enables fast identification and removal of various irregular noise tissue, while simultaneously ensures the complete target regionof interest. Results Our method has been tested by 12 groups of actual clinical cardiac CT sequences, and it segments fast andaccurately, at the same time smooths the edge of region of interest. Conclusion The method is more precise and effective thanother methods like morphology and skeleton extraction, and it provides a well assistance for the diagnosis of clinicians as wellas greatly improves the efficiency and robustness of irregular noise tissue segmentation.