基于改进遗传算法的FISH基因模糊自适应提取

为了快速准确地提取荧光原位杂交（FISH）图像中的荧光染色基因,提高临床医学上诊断病变的效率,针对FISH图像固有的模糊特征,提出了一种基于改进遗传算法的多维多阈值模糊自适应提取算法.首先根据灰度直方图的分布特性,提出相应的自适应窗宽选取算法,确定阈值搜索的范围;然后通过设计的模糊隶属度函数,将图像分割成若干不同的区域;最后采用最大模糊熵准则,并借助优化后的自适应遗传算法,寻找确保基因目标最大信息量的分割阈值.将分割结果与其他几种常用分割算法进行比较,并采用分类概率的数学测量方法定量分析,结果表明：在标准噪声干扰下,该算法的错误划分概率仅为0.042 1,可以更加准确和高效地对荧光基因进行提取.

Fuzzy adaptive extracting of FISH-labeled genes based on improved genetic algorithm

In order to extract the genes in fluorescence in situ hybridization(FISH)-labeled medical cell images precisely and improve diagnostic efficiency,a fuzzy adaptive extracting genes algorithm was implemented based on improved genetic algorithm.In accordance with the inherent distinction of FISH-labeled images,a method of adaptive window size selection was presented.First,a threshold zone was obtained automatically by this method based on the statistic characteristics of the gray level distribution in an image.Then,the FISH-labeled images were divided into several different regions through designing appropriate fuzzy membership functions.Lastly,the criterion of the maximum fuzzy entropy and the optimized genetic algorithm were adopted to find out the best thresholds of extracting genes.The algorithm was evaluated by both real FISH-labeled images and simulated images by the measure method of misclassification error.In all these cases,this algorithm shows a better performance in precision and efficiency than that in the other common algorithms.The obtained results indicate that the probability of misclassification error reaches 0.042 1 when the gray noise is standard noise,which is effective to improve the precision and efficiency of FISH-labeled images segmentation.