感兴趣区域高效提取算法

感兴趣区域在临床医学图像分析中占有重要地位.提出了一种基于单调推进曲线进化的感兴趣区域提取新方法.首先,通过极小化ROI(region of interest)能量函数,推导出区域速度函数项,并与基于边界的速度函数融合,提出融合ROI信息的单调推进Snake模型.ROI信息能够增强曲线深入到对比度低且细窄的区域中的传播能力.其次,提出了多初始化快速推进算法,选择性地种植种子曲线有助于局部区域的生长从而进一步改善分割结果.此外,为提高计算效率,在多尺度空间进行数值求解,其中利用快速解传递方法实现粗一级尺度到细一级尺度解的传递,可以加速收敛.利用医学图像分割实验对该方法进行评估,结果表明:该方法能够快速、精确地提取低对比度和细窄的ROI区域.与现有方法相比,该方法的高效性同时体现在分割结果和计算代价上.

An Efficient Approach to Extraction of Region of Interest

ROI (region of interest) plays an important role in medical image analysis. In this paper, an efficient approach to ROI extraction based on monotonically marching curve evolution is proposed. The improvement is in two aspects: first, a new monotonically marching snake integrating ROI information is presented by minimizing the new defined ROI energy. Due to the region based speed term, the front could even propagate in low contrast and narrow thin areas. Second, a multi-initial fast marching algorithm is developed for numerical implementation, where a multi-initial scheme can perform the selective growth of the front, thus further reduce the front leaking. Furthermore, a multiscale scheme for numerical implementation is adopted, where a fast passing solution method is used for determining the initial solution on the finer scale that greatly reduces the computational cost. The validity of the proposed approach is demonstrated on the medical image ROI extraction. Experimental results show that the approach is efficient both in computational cost and segmentation quality. Low contrast and narrow thin ROI could be efficiently extracted precisely by the approach.