Harmful Algal Bloom Species, Toxin and Total Phytoplankton Identification System (HABStats) for HandHeld, Remotely Networked or Submersible Operation

TECHNICAL ABSTRACT: We propose to develop and commercialize the next generation of low cost, low power, ruggedized, and accurate phytoplankton single cell and whole colony counting, characterization and classification system called HABStats. We will accomplish this using the following three unique and innovative approaches: 1) High-throughput color, Light Field Imaging Flow Cytometry to acquire morphological information (area, major-micro axis, granularity, color pattern, circularity, shape, etc.), combined with 2) Raman and fluorescence spectroscopy to acquire species-specific molecular information on combinations of pigments (B-Carotenes, Chlorophylls, Xanthrophylls, Phycobilliproteins, etc.), triglycerides, and amino acids and proteins. HAB toxins will be characterized and quantified by Raman spectroscopy, and 3) The optical package will be integrated into a flow cytometer-on-a-chip using acoustic focusing to center the stream of cells, and provide a small form factor instrument with a web-enabled processor. Convolutional Deep Neural Networks provide for Deep Learning data integration and artificial taxonomist, and WiFi communications for hand-held, distributed networking, or submersible (1000 m) operations. For Water Quality Monitoring Professionals that need realtime detection and identification of toxic algae and their toxins, CoastalOceanVision’s HABStats imaging and spectroscopic innovation will provide for this need. This proposal directly addresses SUBTOPIC 8.2.1: Portable, Fast, and Intelligent Phytoplankton Species-identifier and Counter.SUMMARY OF ANTICIPATED RESULTS: Completing the work plan in this Phase I proposal will result in a highly detailed design and feasibility report on constructing a prototype flow-through imaging instrument for quantifying and classifying phytoplankton cells, including harmful algae and their toxins in fresh and marine systems. Through building a library of spectral signatures and features extracted from 3-D images of cells, and using Deep Learning techniques like Convolutional Deep Neural Nets, the prototype will have the capability to accurately classify phytoplankton cells and characterize toxins in real-time with high throughput in a hand-held or remote networked package. The commercialization plan is to complete the prototype and its real-world testing and verification in Phase II, followed by Phase III where the instrument will be mass marketed for the detection and management of phytoplankton, both toxic and non-toxic.