Analysis of In-situ Microscopic Organism Behavior in Data Acquired Using a Free-drifting Submersible Holographic Imaging System

This paper introduces a free-drifting submersible digital holography system (holo-sub), shown in fig. 1, and issues associated with processing the massive amount of data involved. Studying behavior of marine micro-organisms in-situ requires the ability to follow them in time and observe their interactions with the local biological, chemical, and physical environment. Holographic imaging allows reconstruction of in-focus images of all planes within a sample volume that has a substantial depth. To facilitate observations of organisms over many frames, thus allowing for analysis of in-situ behavior, the holo-sub is designed to drift with the ambient flow. Two cameras record holograms simultaneously, each at 15 frame/sec. Two optional optical configurations have been implemented to-date. In one configuration, the sample volumes are perpendicular, allowing for 3D resolution of 7.4 mum in all directions. In the second, the first camera maintains a sample volume with a cross-section of 15times15 mm, while the other records holograms of part of the same volume at a higher magnification, enabling us to resolve smaller particles. This system was deployed in the Ria de Pontevedra, Spain in 2005, and in the Bay of Biscay, France in 2006, in collaboration with several European research groups. Over 3 TB of holographic data have been acquired during these deployments. Extracting relevant information from the large holographic dataset requires two important steps: 1 - locating and reconstructing the desired organism, and 2 - processing the reconstructed images to obtain quantitative data on behavior. A series of software tools have been developed to facilitate efficient manual search for organism location, followed by recording of their in-focus images, and tracking them in time. Analysis of reconstructed images to obtain quantitative data begins with classification of behavior, followed by development of image processing routines to perform specific measurements. For example, appendicularians, zooplankton that create gelatinous "houses" with which they filter small particles for feeding are one of the most numerous large organisms that have been observed in our data set. Software has been developed to provide swimming velocity, house shape and size, particle distributions within the house, organism size, and tail flapping frequency.