Introduction to statistical pattern recognition (2nd ed.)
Introduction to statistical pattern recognition (2nd ed.)
Image analysis pipeline for automatic karyotyping
HAIS'12 Proceedings of the 7th international conference on Hybrid Artificial Intelligent Systems - Volume Part II
| Hi-index | 0.00 |
This paper presents a new approach to the classical chromosome classification and identification problem. Our approach maps distinctive features of chromosomes, e.g. length, area, centromere position and band characteristics, into fuzzy logic membership functions. Then fuzzy similarity relations obtained from the membership functions are used to classify and identify the chromosomes. This method has several advantages over classical methods, where usually a prebuilt single-criteria of template chromosomes is used to compare the unknown chromosome as to make a decision about its identity. First the formulation of chromosome characteristics using fuzzy logic better compensates for the ambiguities in the shape or band characteristics of chromosome in the metaphase images, second the use of all the characteristics of the chromosomes produce a more fail-safe method. As a preparatory step to the actual identification process we divide chromosomes according to their fuzzy similarity relation based on length and area into groups. To recover from the situations where a chromosome may be misgrouped because of its disconfirmity to ideal definitions, we refine the grouping of chromosome by applying fuzzy similarity relations which represent the relative centromere positions of chromosomes. Then the band characteristics of each chromosome in a group is correlated with the band characteristics of the chromosomes in the same group of a preprocessed template to obtain identity of the chromosome. The templates used at this step are updated each time when a chromosome is identified, so the system has an adaptive decision algorithm.