Effect of noise and number of dyes in estimating fractional abundance in unmixing of composite fluorescence spectra

Abstract

We report the accuracy of a spectral unmixing algorithm as a function of noise and number of fluorescent probes. The algorithm is based on least-squares minimization with coefficient optimization and is applied to determine the relative fractional abundance of multiple fluorescent dyes from a composite spectrum. Simulation shows that accuracy increases with decreasing noise level and number of fluorescent dyes. An experimental result with 2 to 8% error on three co-localized fluorescent microspheres is presented using the developed approach.