Linear Discriminant Analysis of the wavelet domain features for automatic classification of human chromosomes

Karyotyping is a common method in cytogenetics. Automatic classification of the chromosomes within the microscopic images is the first step in designing an automatic karyotyping system. This is a difficult task especially if the chromosome is highly curved within the image. This paper introduces a new wavelet transform based linear discriminant analysis based feature vector for discriminating both normal and automatically straightened chromosomes in group E. A three layer feed-forward perceptron neural network, which is trained by means of the backpropagation algorithm, is used to classify the input chromosome into one of the three classes in the group E. When tested on a data set of 303 highly curved chromosomes after automatically straightening by a previously reported method by the authors of current article (Roshtkhari and Setarehdan, 2008) an average correct classification rate of 99.3% was obtained.