摘要
目的改善眼底视网膜低对比度血管的检测性能,提高眼部疾病临床辅助诊断的效率和准确率。方法提出了一种基于随机共振机制的眼底视网膜血管多级分割方法。首先用图像信号和附加噪声,获取全局意义下的FHN神经元非线性模型的最佳随机共振响应,对高等级强度信号进行检测;然后将输入信号定义为去除高对比度血管的局部图像,优化参数后,再对低等级强度信号进行检测。最后融合多级随机共振响应,得到眼底视网膜血管的分割结果。结果以DRIVE图像库为例,分别与两位专家手动分割结果进行灵敏度Sn的定量比较,本文方法得到的结果平均值较高,差值分别为0.2007、0.1817。结论本文方法充分利用了噪声对于弱信号检测与分割作用,对低强度等级血管的分割上面优势明显。
Objective To improve the detection performance of low contrast retinal vessels which is helpful to increase the efficiency and accuracy of clinical auxiliary diagnosis of ocular diseases. Methods A multilevel segmentation method for retinal vessels based on stochastic resonance was presented in this paper. Firstly,the optimal stochastic resonance response of FHN neuron nonlinear model in the global sense was obtained by image signal and additional noise,and the high-level signal was detected. Then the input signal was defined as the local image of high contrast blood vessel. After optimizing the parameters,the low intensity signal was detected. Finally,the multilevel stochastic resonance response was integrated to obtain the result of retinal blood vessel segmentation. Results Taking DRIVE image library as an example,the sensitivity Sn of the method was quantitatively compared with the manual segmentation results of two experts. The average value of the results obtained by this method was higher,and the difference was 0. 2007 and 0. 1817,respectively. Conclusion The new method makes full use of the effect of noise on the detection and segmentation of weak signals,and has obvious advantages on the segmentation of low intensity vessels.
作者
杜宇华
范影乐
武薇
Du Yuhua;Fan Yingle;Wu Wei(Laboratory of Pattern Recognition and Image Processing,Hangzhou DianZi University,Hangzhou Zhejiang 310018,China)
出处
《航天医学与医学工程》
CAS
CSCD
北大核心
2018年第6期650-658,共9页
Space Medicine & Medical Engineering
基金
国家自然科学基金(61501154)
