一种多尺度圆形滤波器在肺结节增强中的应用

肺部CT图像的结节检测由于受到噪声和肺部区域中气管、血管的干扰,一直是医学辅助诊断领域中的一个难点。针对这一问题,提出了一种基于多尺度的各向异性平滑方程的圆形滤波器方法。该方法首先利用各向异性平滑方程与高斯圆模型的Hessian矩阵推导出多尺度的圆形平滑方程,利用该方程对结节图像进行多尺度平滑;然后分析了高斯圆模型Hessian矩阵特征值的特点,建立了圆形增器滤波函数。最后,利用该滤波函数对多尺度平滑后的结节图像进行圆形增强滤波。实验证明该方法能够有效地抑止非圆形状的干扰,得到较好的结节增强图像,为后续的结节特征提取与分类奠定了基础。

Algorithm for Circle Filter Based Anisotropy Smoothing Equations Application for Enhancement of Pulmonary Nodules

The detection of lung nodules is a difficult problem in the field of medical aided diagnosis, due to the noise and the disturbance of blood vessels and tracheas. According to this problem, a Circle filter algorithm was brought forward based on the anisotropy smoothing equations. A circle smoothing equation was deduced by means of the anisotropy smoothing equations and Hessian matrix of the circle in the Gaussian model. By analyzing the characters of eigenvalues of the Hessian matrix, a function for the circle enhancement filtering was proposed. The circle enhancement filter was then used to enhance the nodule image, which was smoothed in different scales. The experiment result shows that this algorithm has effectively suppressed the non-circle disturbance; hence enhanced images are obtained, which will be beneficial for the feature extraction and classification of nodules.