Improving phytoplankton abundance estimation accuracy for autonomous microscopic imaging systems

Abstract

A novel method to accurately estimate phytoplankton abundance is proposed for an autonomous microscopic imaging system (AMIS) herein. To this end, a fast fluorescence detection module is developed and added to an imaging in-flow cytometer to record the fluorescence and side-scattered signals of individual phytoplankton particles, including of those that cannot be photographed by the AMIS. Image information and the coupling relationship between the fluorescence and side-scattered signals are used to accurately detect and estimate the phytoplankton counts in water samples. The performance of the proposed estimation method is evaluated on water samples containing Alexandrium tamarense, Chattonella marina, and Scrippsiella trochoidea. The abundance estimation accuracies for these species are found to be better than 95%, 97%, and 93%, respectively, when compared to results obtained using counting chambers. The performance of the method is further evaluated by mixing the collected data of the three phytoplankton species and classifying them based on fluorescence and side-scattered signals only, assuming that these species are included in the image data but not photographed individually. The overall estimation accuracy based on this complex matrix of the three species is found to be 95.3%. These results demonstrate the suitability and practicality of the proposed method for accurately evaluating phytoplankton abundance in water. The algorithm used in this study can be a reference for other imaging in-flow cytometers.