Classification of chromosomes using a combination of neural networks

In developing computer vision systems for analyzing chromosome images, a central task is the classification of the 46 chromosomes into 24 groups. A combination of multilayer-perceptrons for classifying isolated chromosomes is described. It is demonstrated that these perform as well as, or significantly better than, a well-developed statistical classifier. A method is suggested for using a competitive network to take advantage of constraints on the assignment of chromosomes to groups as a means of improving the classification rate.<>