Spot Counting for Automated Analysis of Unmixed Hyper-Spectral M-FISH Images

Fluorescence microscopy plays an important role in the classification of cancerous tissue. Fluorescent in situ hybridization (FISH) is a common method for marking different cell components (e.g. nucleus, cytoplasm, proteins) as well as specific DNA positions or whole DNA sequences. The FISH method/variant used in this paper facilitates a special combination of filters (DAPI/Orange/Green) and corresponding fluorescent dyes to discriminate between the different parts of a cell. FITC marks chromosome 17, Spectrum Orange R © marks HER-2/neu receptors and DAPI is used to counterstain the cell nucleus. Upon excitation, each marked chromosome emits a fluorescent signal (spot). A common problem with multicolor FISH (M-FISH) preparations is the crosstalk between the channels caused by the overlap of the emission spectra of the different fluorophores. This crosstalk is one of the reasons that the evaluation and classification of M-FISH preparations is difficult and requires experienced experts. The crosstalk cannot be resolved on the filter level (excitation/emission), and not by specialized fluorophores (which have different emission spectra). However, the crosstalk can be eliminated if spectral imaging techniques are used to acquire hyperspectral image data of M-FISH preparations and employ spectral unmixing methods to ”algorithmically reduce” the spectral overlap of the emission spectra. Spectral imaging is the combination of computer vision and spectroscopy. And due to the fact that every object of interest consists of more than one pixel, every pixel is dependent on its neighboring pixels. Thus the spatial context of the image contains useful information for a classification and increase the sensitivity and specificity of a spectral classification. Keywords—spectral imaging (SI), fluorescence microscopy, fluorescent in situ hybridization (FISH), contextual classification.