一种新的舌癌图像快速自动分割方法

目的:研究一种新的舌癌图像自动分割算法以实现对舌癌肿瘤的快速准确分割。方法:通过引入一种基于局部均方差的自适应尺度算子实现演化曲线在演化过程中的自动调整，从而更高效率地向真实目标边界运动，并且克服舌癌肿瘤图像中目标边界不清和图像灰度不均匀等不良因素带来的影响。此外，为加快曲线的收敛速度，本文提出了一种新的能量项评估演化曲线轮廓内部和轮廓外部区域灰度的分布差异，以此引导曲线自适应地调整演化速度，减少完成分割任务所需的迭代次数。结果:使用本方法对22幅舌癌肿瘤MRI图像进行分割，分割结果与真实结果之间的重叠率Dice值为0.82，豪斯多夫距离HD值为1.732 mm。结论:将本文算法与其它现有的几种活动轮廓模型进行定性和定量对比分析，实验结果表明本文算法在对细节及弱边缘灰度的处理上表现更加优异，可用于舌癌肿瘤的精确分割，为临床分析提供辅助信息。

A new and fast method for automatic tongue cancer image segmentation

Abstract: Objective To develop a novel method for automatic tongue cancer image segmentation, thereby realizing the quick and accurate segmentation of tongue tumor. Methods An adaptive scaling operator based on local mean-square error was introduced to realize the automatic adjustment of the evolution curve during the evolution process, thus making the curve move more efficiently to the real target boundary and overcoming the effects of bad factors such as unclear target boundary, edge blur and uneven gray level in tongue cancer image. In addition, in order to accelerate the convergence speed of the curve, a new energy term was proposed to evaluate the intensity distribution difference between the inner and outer regions of the evolution curve contour, so as to guide the curve to adjust the evolution speed adaptively and reduce the number of iterations required to complete the segmentation task. Results Twenty-two MRI images of tongue cancer were segmented by the proposed method. The overlap ratio (Dice value) between the segmented results and the real results was 0.82, and the Hausdorff distance (HD) was 1.732 mm. Conclusion The qualitative and quantitative comparisons between the proposed method and other existing active contour models reveals that the proposed algorithm is advantageous in the processing of details and weak edge gray level, and that it can be used for accurate segmentation of tongue tumor, thereby providing auxiliary information for clinical analysis.