LSU Logo LSU Logo
Login to myLSUFriday, 16-Nov-2018 03:44:01 CST
The lengthy processing time for enumerating preserved zooplankton samples has been one of the roadblocks to complete analysis of time-series data such as the SEAMAP collection. The time and expense of paying trained plankton taxonomists, as well as the paucity of taxonomists, has limited the number of samples that can be analyzed. Thanks to the dedication of the personnel involved in the SEAMAP program over the past 27 years, the samples that were collected have been carefully preserved and archived and with the development of new technologies, we are now able to analyze the vast numbers of samples that have been in storage as well as those that are being currently collected.

We are utilizing the latest sample imaging techniques to create a digital record of each sample that can be shared with the larger scientific community and the latest semi-automated analysis techniques to process the surviving SEAMAP zooplankton samples. We are using a ZOOSCAN (Grosjean et al. 2004), an instrument that uses a modified scanner sensor with a custom-built lighting system and a water-tight scanning chamber into which zooplankton samples can be poured, digitized at high-resolution, and recovered without damage (Fig. 5). The contents of these images then can be identified, enumerated, and measured by a human working with specialized software. The system was designed by Dr. Gabriel Gorsky, a renowned zooplanktologist working in collaboration with engineers and software developers. It is manufactured, marketed, and supported by the French engineering firm RECIF Technology.

This is the Zooscan website:

Samples of plankton are poured on to a scanning tray and individual zooplankton are arranged to distribute the sample across the tray. With a computer interface, the user then scans the sample at 2400 dots per inch (dpi) at 16 bit grayscale resolution and saves the contents to a TIFF raster file (Fig 6). The scanned image is calibrated using a previously collected background image (a blank image).

Zooscan scanner
Figure 5. ZOOSCAN is a specially-designed waterproofed scanner system that can rapidly digitize the contents of preserved or freshly-collected plankton samples.

Digitized images of zooplankton (Fig. 6) from the ZOOSCAN have the advantage of providing a clear record of the contents of each sample in a form that can be stored electronically and archived on removable media such as DVDs. The entire sample can be digitized in a series of scanned images or the sample can be size-fractionated and then subdivided (split) to obtain representative aliquots of the contents. Each ZOOSCAN is calibrated in terms of its grayscale values and image dimensions. ZOOSCAN images analyzed with one system will produce exactly the same output information as another ZOOSCAN.

Software available with the ZOOSCAN provides a means of locating each object in the image. These objects (termed regions of interest: ROIs) are cropped from the image, measured and written to disk as a series of small image files. Once this process is complete, the image file and resulting data on the individual particles are uploaded to the database at LSU.

Zooscan scan sample
Figure 6. An example scan of a subsample of the zooplankton from SEAMAP sample SIPAC 30017. The upper image shows a greatly reduced image of the complete scanned area. Certain individuals have been highlighted and shown at higher magnification in the lower section of the scan. A, B: chaetognaths; C: decapod; D: fish larva (likely a hatchetfish); E: pteropod; F: decapod: G, H: calanoid copepods; I: pteropod; J: ostracod; K: foraminiferan; L: ostracod; M: pteropod; N: larvacean.

In some cases, additional analysis is conducted in which the zooplankton are identified using a combination of automatic computer identification and validation by a trained technician. A learning set is created by the technician who sorts individual images of zooplankton into different categories (Fig. Z—a possible figure showing the computer interface for creating a learning set). Then, the learning set is used to predict the identity of the zooplankton in each sample. The computer identifications are generally 80% correct, and so validation by a trained technician is essential. Still, there are significant time savings by using the computer-based identification techniques rather than sorting the individual zooplankton images manually.