Method Research For Automatic Chromosome Recognition Using Dynamical Threshold Algorithm And Characteristic Bands

Chromosome recognition is a problem of consequence in cytogenetics, and is widely used in chromosome syndrome diagnosis. The traditional chromosome recognition is preformed manually, which is tedious and time-consuming with relative low recognition rate. Since late 1960s, Automatic chromosome recognition is studied worldwide, and many such systems were launched.Automatic chromosome recognition is the hotspot in pattern recognition, machine vision and artificial intelligent, and its achievements have the great value in cytobiology, genetics and clinical medicines. So people are attracted in this domain all the time. Although the automatic chromosome recognition is adopted, high quality picture and / or human intervention is required, otherwise, veracity would drop. So the recognition systems nowadays are not the automatic ones in word and deed.Statistical recognition is often employed in the early stage in chromosome recognition, with the development of physiology, neurobiology, psychology, Artificial Neural Networks is adopted, and recently Fuzzy Theory is introduced. Recognition rate improves a lot using all the methods, although hasn't turn up trumps.Studying the automatic recognition system, the tough work mainly includes: (1) Single chromosome extraction is not always done preciously due to the inhomogenous illumination. (2) Chromosome length and centromere index are traditional classification foundation. But often they are not clear enough, due to the staining process and image capture and pre-process.In view of the above problems, Dynamical Threshold Algorithm is introduced to recognize edges, and Characteristic Bands are put forward to provide new chromosome features. Meanwhile Skeleton Extraction is improved also. Comparison with traditional methods, recognition rate improves a lot, especially for the chromosome with close length and CI.