Abstract
Brain tissue contains autofluorescing elements that potentially impede accurate identification of neurons when visualized with fluorescent microscopy. Age-related accumulation of molecules with autofluorescent properties, such as lipofuscin, can possess spectral profiles that invade the typical emission range of fluorophores commonly utilized in fluorescent microscopy. The traditional method for accounting for this native fluorescence is to apply lipophilic dyes that are able to sequester these unwanted signals. While effective, such dyes can present a range of problems including the obstruction of fluorescent probe emissions. The present study utilizes aged primate midbrain tissue stained for tyrosine hydroxylase and calbindin to investigate an image processing approach for removing autofluorescence utilizing spectral imaging and linear unmixing. This technique is then compared against the traditional, dye-based autofluorescence sequestration method using Sudan Black B (SBB). Spectral imaging and linear unmixing yielded significantly higher cell numbers than SBB treatment. This finding suggests that computational approaches for removing autofluorescence in neural tissue are both viable and preferential to dye-based approaches for estimation of cell body numbers.