基于稀疏表示和随机游走的磨玻璃型肺结节分割

Segmentation of Ground Glass Opacity Pulmonary Nodules With Sparse Representation and Random Walk

肺结节是早期肺癌在影像学上的表现形式.磨玻璃型(Ground glass opacity,GGO)肺结节被认为是恶变可能性最大的一类结节之一.针对GGO结节边缘模糊、大小各异、形状不规则和灰度不均匀等造成分割准确率低问题,本文提出了一种基于稀疏表示和随机游走模型的分割算法.首先,利用测地距离和局部搜索策略,自动地选取了种子点.其次,联合8-邻域和稀疏表示的K-最近邻算法建立了新的图,避免了噪声的干扰.结合灰度、纹理、空间距离和稀疏系数构建了新的加权矩阵.最后,将标签限制项引入到随机游走的能量函数中.该算法分割准确性较高,鲁棒性较强.

Pulmonary nodules are the radiographic manifestation of lung cancer in early stages. Ground glass opacity（GGO） pulmonary nodules are considered to be one of the most likely nodules of malignancy. To address the low accuracy segmentation problem caused by blurred boundaries, different sizes, irregular shapes and inhomogeneous intensities of GGO nodules, a segmentation algorithm with the sparse representation and random walk model is proposed. Firstly, the geodesic distance and local search strategy are introduced to automatically select seeds. Secondly, 8-neighbor and sparse representation K-nearest neighbor algorithm are combined to build a new graph which avoids the interference of image noise. To construct a new weighted matrix, intensity, texture, spatial distance and sparse coefficients are incorporated.Finally, a label constraint term is added to the energy function of random walker. The proposed algorithm can obtain a high accuracy and strong robustness.