Identification of phytoplankton from flow cytometry data by using radial basis function neural networks

Area of Science:

• * Algorithmic biology and computational fluid dynamics.
• * Machine learning applications in ecological monitoring.

Background:

• * Accurate identification of phytoplankton is crucial for aquatic ecosystem monitoring and research.
• * Traditional methods for phytoplankton identification can be time-consuming and require specialized expertise.
• * Flow cytometry offers high-throughput data acquisition but requires sophisticated analysis for species-level identification.

Purpose of the Study:

• * To evaluate the efficacy of Gaussian radial basis function (RBF) networks for identifying phytoplankton species.
• * To analyze the impact of network parameters on the optimization of RBF networks for this task.
• * To assess the performance of RBF networks in recognizing both known and novel phytoplankton species.

Main Methods:

• * Utilized an artificial neural network, specifically the Gaussian radial basis function (RBF) network.
• * Employed 11-dimensional flow cytometric data generated by the European Optical Plankton Analyser instrument.
• * Trained and tested the RBF network on data from 34 marine and freshwater phytoplankton species.

Main Results:

• * Optimized RBF networks achieved an overall success rate of 91.5% in identifying phytoplankton species.
• * Analyzed the relative importance of each flow cytometric parameter in discriminating between species.
• * Investigated the network's response to data from novel phytoplankton species not included in the training set.

Conclusions:

• * Gaussian RBF networks provide a highly successful and efficient method for automated phytoplankton identification.
• * The study demonstrates the potential of machine learning in advancing aquatic microbial ecology.
• * RBF networks show promise for identifying unknown phytoplankton species, aiding ecological surveillance.