CyanoSCOPE: an open-source, deep-learning approach to automate cyanobacteria identification and enumeration from microscopy imaging

Citizen scientists are crucial in monitoring cyanobacteria harmful algal blooms (cHABs); however, the success of these programs hinges on certain requirements. To ensure user satisfaction and maintain interest, the process of identifying these blooms needs to be rapid, reliable and easy to use. To address this issue, CyanoSCOPE, a deep-learning imaging tool has been developed in the R programming language using Keras and UNet neural network modeling. This tool can identify single-cell and filamentous cyanobacteria from light microscopy images. Automated quantification tools are built from the cellcount library, using tools and functions from EBImage. The current version of CyanoSCOPE can identify Microcystis, Dolichospermum, Anabaena, and Sphaerospermopsis genera. This tool also has built-in identifiers for common freshwater diatoms, chlorophytes, chrysophytes, and cryptophytes; these additional features help demonstrate overall accuracy, precision, and efficiency. An interactive user interface enables users to upload images and apply models seamlessly, and generates identifier data and bloom density metrics. The applications of this tool range from cyanobacteria culture maintenance to endless possibilities within the citizen-science sector where citizens can easily locate, identify and monitor cHABs as they progress. It also eliminates the time-consuming bottleneck of cell identification by microscopy, which can further delay a cHAB response and reduces potential identification discrepancies between taxonomers.

Primary Presenter: Tyler Harman, CSS, Inc (Contracted to NOAA NCCOS) (tyler.harman@noaa.gov)

Authors:

D. Ransom Hardison, National Oceanic and Atmospheric Administration: National Centers for Coastal Ocean Science  (rance.hardison@noaa.gov)

William Holland, National Oceanic and Atmospheric Administration: National Centers for Coastal Ocean Science  (chris.holland@noaa.gov)

Kaytee Pokrzywinski, National Oceanic and Atmospheric Administration: National Centers for Coastal Ocean Science  (kaytee.pokrzywinski@noaa.gov)