M-FISH image registration and classification

Multiplex or multi-color fluorescence in situ hybridization (M-FISH) imaging is a recently developed cytogenetic technique for cancer diagnosis and research on genetic disorders. By simultaneously viewing the multiple-labeled specimens in different color channels, M-FISH facilitates the detection of subtle chromosomal aberrations. This technique largely depends on the accurate pixel classification (color karyotyping). We propose a Bayesian classifier for multispectral pixel classification. Due to color aberration and other source of errors, the misalignment among the different fluor images becomes inevitable, resulting in misclassified pixels. A multiresolution registration algorithm is introduced for this purpose, which is based on wavelets and spline approximations. The effect of the registration on subsequent classification was evaluated on the ADIR M-FISH database. It indicates that the proposed registration technique leads to increased pixel classification rate, translating into improved accuracy in identifying subtle DNA rearrangements.