Chromosome image classification using a two-step probabilistic neural network

Chromosome image analysis is composed of image preparation, image analysis, and image diagnosis. General procedure of chromosome image analysis includes of image preprocessing in the first step, image segmentation, feature extraction, and image classification in the last step. This paper presents the preliminary results that use probabilistic neural network to classify chromosome image into 24 classes. Features of chromosome which were used in this paper are area, perimeter, band’s area, singular value decomposition, and band profile. Chromosome images were grouped in two steps by probabilistic neural network. Six groups and twenty four groups are in the first and the second step, respectively. The result from the second step is twenty four chromosome classes. Density profile sampled at 10, 30, 50 and 80 were tested. The best classification result of female is 68.19% when density profile at 30 samples was used, and that of male is 61.30% when density profile at 50 samples was used.