基于自适应RSF模型的水平集桡骨分割方法研究

骨块的准确提取是骨龄自动识别过程中重要的一步,桡骨是骨龄评分中一块重要的较难自动分割的骨骼,故桡骨的自动分割具有重要研究意义。文中提出了一种根据闭合曲线内外图像信息熵自动调整权值的改进水平集方法。鉴于X线影像对比度低,文中选用基于区域可变的拟合模型的水平集算法来进行分割,该算法具有不需预知待检图像拓扑模型,且在复杂拓扑情况下进行图像分割效果较好的优点,但该算法需对其中的权值进行手动调整才能实现准确分割;为了进一步实现自动分割,改进的算法根据闭合曲线内外图像信息熵自动调整权值,从而实现自适应桡骨分割。通过与原水平集方法的分割结果进行对比实验证明,改进方法能有效的实现桡骨分割,且有很强的自适应性。

Research on the Segmentation of Radius Based on Level Set Method of Adaptive RSF Model

The correct extraction of bone is a very important step in bone age assessment process.Radius is a vital bone that is hard to segment automatically in bone scoring,so it is significant to implement auto-segmentation of radius.In this paper,an improved level set algorithm in which the weighted values are tuned adaptively is proposed in term of entropy inside and outside the contour.Because of the low abstract of the X-ray image,a region-scalable fitting model is used to make the segmentation.This model has good performance of which the topology model of the processed image needn＇t to be foreseen,moreover it could leave a right result in complicated topology,but in order to segment precisely,the weighted coefficients should be adjusted manually.To achieve further auto-segmentation,the weighted coefficients are updated with entropy inside and outside the contour,and so as to realize the adaptive segmentation of radius.Compared with the original method,experiments show that the algorithm is effective and adaptive.