Research Of Automatic Chromosome Karyotyping System

Chromosome Karyotyping, the classical method in the technology of chromosome genetic analysis, is one of the important means in genetic research and supplementary clinical diagnosis. And it is then key index to analyze chromosome translocation or deficiency,and diagnosis of a variety of genetic diseases. With the development of pattern recognition ,and image processing methods begin widely used in medicine ;automatic chromosome analysis becomes possible .This paper mainly researches on numerical analysis and processing method of biologically medical image on the background of image processing , neural network and pattern recognition of human Chromosome images. Some important technologies and core algorithms used in the system are discussed in detail, including the algorithms of image pre-processing, feature measurement, selection and classification. The main results and achievements are showed as follows:1. In the procedure of chromosome image preprocessing, At first, we use Median Filter smoothing the image, then calculate threshold by Iteration Method, the image which is dealt with by the threshold has good detached character.2. After the processing of chromosome image, using the method of tab, we can finish automatic division of chromosome image and get the single chromosomes, then describing the boundary each chromosome. The experiment results show that image pre-processing procedure of this system proposed in this article can effectively formed a good foundation for extracting Chromosome characteristics.3. This paper relates the method of extracting Chromosome medial-axis-line, calculating the length of Chromosome, calculating the centromeric index and density profile from human Chromosome images.4. Extraction of medial axis of chromosomes is also one of the most important parts of the system. We propose a method for medial axis extraction combining the algorithms of thinning, branch removing, curve fitting and optimized curve extending. By the method, satisfactory medial axis can be extracted. The method is the base of calculating the length of Chromosome, calculating the centromeric index and density profile.5. We present a algorithm of finding the centromeres that combining the curvatures of contours and the crossing detection of chromosomes. With the algorithm, the accuracy of contromere detection is increased.6. This text realizes Multi-stage Classifier .In the first layer, according to neural network pattern recognition based on BP algorithm, Chromosomes are classfied as A,B,C,D,E,F,G,X,Y . In the second layer, we use k-nearest neighbor rule realizes to classify the chromosomes. In experiment, we select relative length, arm ratio, centromer index and density profile to classify the chromosomes.The experiment results indicate that, if the chromosomes on then slide isolation each other, and there are no overlapping chromosomes , the advantage of this system is its higher automation and its better accuracy for classification.