Use of fluorescence information for automated phytoplankton investigation by image analysis

Abstract

Automated identification and quantification of algae in microscopic images is a tool that allows high taxonomic resolution with reasonable technical efforts. However, in samples containing various non-algal objects, this is still not a satisfactorily solved problem. We show that autofluorescence information improves discrimination of algae from non-algal objects as well as phycoerythrin (PE) containing algae from others. We analyse the stability of the autofluorescence to estimate its constraints. Cold and dark storage of glutaraldehyde fixed samples maintains autofluorescence sufficiently for 3 weeks. Under repeated excitations, chlorophyll a (Chl a) or PE autofluorescence show an exponential decrease followed by an intermediate maximum. A peak also occurs in emission wavelength ranges without chlorophyll and PE fluorescence. The unspecific autofluorescence causing the peaks is at least partly identical with the blue-green fluorescence (BGF) in plant cells. BGF interferes with identification of algae, thus correction of pigment autofluorescence with such unspecific fluorescence allows a more reliable algal discrimination procedure. A classification scheme for discrimination of Chl a and PE-containing algae shows a high performance in a test with natural samples. Integration of fluorescence and bright-field image information provides a powerful tool for phytoplankton analysis in complex samples. © The Author 2008. Published by Oxford University Press. All rights reserved.