AUTOMATED ANALYSIS OF MULTI-SPECTRAL M-FISH IMAGES FOR BREAST CARCINOMA STAGING

摘要

M-FISH images are difficult to interpret because the emissionspectra of fluorochrome marked DNA probes overlapwith each other and with the tissue’s intrinsic autofluorescence.Multi-spectral imaging is used to take spatiallyand spectrally resolved images of M-FISH samplesand spectral unmixing is applied to resolve the crosstalk inthe emission spectra. The unmixing is realized by a combinationof: Alternating Least Squares (ALS) and VARIMAX.Spot counting algorithms use the high contrast falsecolor images to automatically determine the HER-2eu toCEP 17 ratio.