Premium
Improving a semi‐automated classification technique for bivalve larvae: Automated image acquisition and measures of quality control
Author(s) -
Goodwin J. D.,
North E. W.,
Mitchell I. D.,
Thompson C. M.,
McFadden H. R.
Publication year - 2016
Publication title -
limnology and oceanography: methods
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.898
H-Index - 72
ISSN - 1541-5856
DOI - 10.1002/lom3.10123
Subject(s) - crassostrea , magnification , artificial intelligence , image processing , computer science , computer vision , data acquisition , image quality , biology , pattern recognition (psychology) , ecology , image (mathematics) , oyster , operating system
Bivalve larvae are small (50–400 μ m) and difficult to identify using standard microscopy, thus limiting inferences from samples collected in the field. With the advent of ShellBi, an image analysis technique, accurate identification of bivalve larvae is now possible but rapid image acquisition and processing remains a challenge. The objectives of this research were to (1) develop a benchtop automated image acquisition system for use with ShellBi, (2) evaluate the system, and (3) create a protocol that would maintain high classification accuracies for larvae of the eastern oyster, Crassostrea virginica . The automated system decreased image acquisition time from 2–13 h to 46 min per slide and resulted in the highest classification accuracies at the lowest tested magnification (7X) and shortest image acquisition time (46 min). Quality control tests indicated that classification accuracies were sensitive to camera and light source settings and that measuring changes in light source and color channel intensities over time was an important part of quality control during routine operations. Validation experiments indicated that under proper settings, automated image acquisition coupled with ShellBi could rapidly classify C. virginica larvae with high accuracies (80–93%). Results suggest that this automated image acquisition system coupled with ShellBi can be used to rapidly image plankton samples and classify C. virginica larvae allowing for expanded capability to understand bivalve larval ecology in the field. Additionally, the automated system has application for rapidly imaging other planktonic organisms at high magnification.