Chromosome Analysis Based On Fuzzy Theory And Neural Artificial Network

The analysis and recognition of chromosomes are very important problem in cytogenetics, and it is widely used for diagnosis of genetic disease. The traditional recognition of chromosome is performed by hand .So the process of recognition is tedious and will cost much manpower.The recognition method of chromosome adapts statistical recognition. Later application of neural network for the recognition increases notably the recognition rate. But it never reach our expect. We think that Main difficulties lie in: (1) centrometric index is a main feature of recognition in former system of chromosome recognition. But because stained with Giemasa and affected by image processing, centromere of chromosome isn't very distinct. It is difficult to find the place of centromere on chromosome precisely. So centromeric index can't get an accurate value. (2) It is difficult to describe band information of chromosome, because band is indistinct, edge between bright band and dark band is not distinct, two bands overlap, and shape of band is irregular.(3) Technique of separation of chromosome require to improve, especially separation of crossing chromosome and overlapping chromosome.In view of above difficulties, Fuzzy theory is applied to chromosome recognition. Combining fuzzy theory with neural network, we present a fuzzy neural network and used it as classifier for chromosome. The experiment results show that recognition rate improve a lot comparing to the traditional BP network. Moreover, we make effort to do some work about individual chromosome extraction and individual chromosome image processing.