基于改进图割的超声瞳孔图像分割与直径测量

Pupil Ultrasound Images Segmentation and Diameter Measurements Based on Improved Graph Cuts

传统的瞳孔直径测量是通过医生手工标定,对于眼外伤和丧失意识的患者测量不方便。针对瞳孔直径测量的人工交互量大且测量鲁棒性不强的问题,采用图割算法分割瞳孔超声图像并测量瞳孔直径。对传统图割算法进行两个方面的改进,采用自适应阈值的区域生长代替人为种子点选取,在保证分割效果的基础上减少了图割的交互量;在能量函数的数据项部分增加图像的梯度信息,减少了原始算法分割结果中出现的小区域,增强了对弱边缘的分割。最后,对采集到的超声瞳孔图像进行自动分割、自动测量瞳孔直径,可以得到患者瞳孔的直径动态变化,给临床诊断提供依据。为了验证算法的有效性,对10位患者的动态瞳孔超声图像进行基于改进图割的瞳孔直径测量,并与医生的手动测量结果对比。结果表明,本方法的结果与医生手动测量结果的绝对误差小于0.2 mm,相关系数不小于0.83。通过改进图割算法,改善了分割效果,实现了超声瞳孔动态图像的自动直径测量,并可有效代替瞳孔直径的人工测量,减少人工交互量。

The traditional measurement of pupil diameter is manually measured,but for patients who have ocular trauma or loss of consciousness this method is inconvenience. Aiming to solve the problem of large manual interaction in pupil diameter measurement and the weak measuring robustness,we used improved graph cut segmentation algorithm to segment pupil ultrasound images and measure pupil diameters. In this paper,we improved the traditional graph cut algorithm in two aspects. One is using adaptive threshold region growing to take place of manually seeds selection,which ensures the segmentation results while reducing the amount of manual interaction. The other one is increasing gradient information of the image into data entry portion of the energy function,which reduces the small area in the segmentation results and enhances the weak edges of segmentation. The proposed method realized automatic segmentation of image and automatic measurement of the pupil diameter. By employing the method we acquired dynamic changes of patient pupil diameter and provide a basis for clinical diagnosis. To verify the validity of the algorithm,we used the method to measure the diameter of ten patients ＇ dynamic pupil ultrasound images,and compared the results with that obtained by manual measurements. It is shown that the absolute error is less than 0. 2 mm,the correlation coefficient is at least0. 83. In conclusion,the modified graph-cut algorithm improved segmentation results,achieved automatical measurements of pupil diameters using dynamic ultrasound pupil images,and can be expected to substitutemanual measurements and reduce the amount of human interaction.