Detection of chromosomal abnormalities with multi-color fluorescence in situ hybridization (M-FISH) imaging and multi-spectral wavelet analysis

Multicolor karyotyping technologies, including both multi-color fluorescence in situ hybridization (M-FISH) [l] and spectral karyotyping (SKY) [2], are recently developed molecular cytogenetic techniques for rapid visualization of genornic aberrations at sub-cellular level. Multi-color karyotyping techniques are based on multi-spectral microscopic imaging in combination with high resolution molecular probes developed in genomics, which have been successhlly used in postnatal and cancer cytogenetics. Despite their promise, the current techniques are not reliable for clinical use [3]. Routine clinical application of these techniques for molecular cytogenetic diagnosis has been hampered by several technical limitations. A primary problem is the pixel by pixel classification of chromosomes from the multi-color image representations. In this work, we propose a new wavelet transform for multi-color image representation, which can effectively reduce the redundancy between multi-color images such that the classification accuracy can be improved. The algorithm was validated by testing on a database containing chromosome images obtained from a variety of M-FISH probes.