Chromosome pairing for karyotyping purposes using mutual information

Cytogenetics is the preferred tool in the diagnosis of genetic diseases such as leukemia and detection of aquired chromosomal abnormalities, such as translocations, deletions, monosomies or trisomies, etc. The karyotyping is a set of procedures, in the scope of the cytogenetics, that produces a visual representation of the 46 chromosomes, paired and arranged in decreasing order of size, observed during the metaphase step of the cellular division (meiosis). The pairing is the procedure in the karyotyping process where the homologous chromosomes are paired according to dimensional, morphological and textural similarity criteria. This process is time consuming and is usually performed manually by experts. An automatic pairing algorithm is still an open problem. In this paper we present new contributions to solve the automatic pairing problem in the scope of the karyotyping process for leukemia diagnostic purposes. Besides the traditional features used to compute the similarity between chromosomes, such as, normalized area, ellipsis axis length and banding profiles, we introduce the Mutual Information (MI) measure to assess the textural similarity between two chromosomes. A supervised linear classifier is trained to combine the different features computed from each pair, aiming at the correct pairing (as given by experts). The resulting classifier is then employed, together with a combinatorial optimization algorithm based on A*, to compute the pairing for any given image. Simulations using real images, obtained with a Leicatrade Optical Microscope DM 2500, were performed. These images were manually paired by experts and used as a ground truth for the pairing process to assess the performance of the proposed classifier. Furthermore, qualitative comparisons with the results obtained with a Leicatrade CW 4000 Karyo software were also performed.