Summary

The Zooscan system is used to digitally image, classify, and measure preserved mesozooplankton organisms.

Methods

1. Principle

The Zooscan digital imaging system allows mesozooplankton to be imaged, classified, and measured digitally. This digital method provides several advantages to traditional counting using microscopic analysis. The Zooscan provides a digital archive of samples, making it feasible to store the images long term and to share the images with others. The digital images are measured automatically, removing human error and bias from the process.

The study region selected is off Southern California (CalCOFI line 80, shore to station 100; line 93, shore to station 120; and lines 83 and 87, shore to station 70, inclusive)

All samples have been collected with a 50-cm diameter PRPOOS net from July 2005 to present. For details on net sampling protocols at sea, see: PRPOOS net deployment.

2. Sample Size Fractionation

2.1

The sample is poured gently into stacked sieves with meshes of 1000 µm and 200 µm.

2.2

The samples are rinsed through the sieves using filtered sea water.

2.3

The organisms are rinsed from the sieves into clean jars and the volumes of the fractionated samples brought to 400 ml using filtered sea water.

3. Begin a new Zooscan project

3.1

Open Zooprocess software.

3.2

Choose the "Create NEW project" process.

3.3

Select the C:\ drive as the destination for the project.

3.4

Enter the project name. It can include only letters (lower case), numbers, and _s.

3.5

Make sure that the "Process also new measurements" box is checked.

3.6

Select the "natural" image process method.

3.7

Modify the config file in the project folder with the specifications for your machine.

4. Background Scans

4.1

Before the first scan is performed each day, background scans (scans of the Zooscan scanning surface with no organisms) need to be performed. These scans serve two purposes: they will be used to normalize the background after zooplankton samples are scanned and they can be used to track changes in the Zooscan hardware through time.

4.2

Turn on the Zooscan.

4.3

Clean the scanning area of the Zooscan using glass cleaner and paper towels.

4.4

Fill the scanning area with deionized water.

4.5

Make sure that the narrow frame is clean (no water marks, etc.) If it needs to be cleaned, do so using glass cleaner and paper towels.

4.6

Place the narrow Zooscan frame on the scanning area. Align the blue dots so that the dot on the Zooscan matches up with the blue dot on the narrow frame. The frame should be flush against the left edge of the scanning area.

4.7

Open Zooprocess software.

4.8

Select the project from the project list.

4.9

Choose the "SCAN (CONVERT) Background Image" process.

4.10

Press the light initialization button on the Zooscan.

4.11

Open VueScan software.

4.12

Push the preview button in VueScan to preview the background image. Check the preview image to make sure that there are no lines or spots on the images.

4.13

Wait at least 30 seconds after pressing the preview button, and then push the scan button in VueScan.

4.14

Push the OK button on the Zooprocess dialog box.

4.15

After the scan has completed, wait at least 30 seconds before pushing scan a second time.

4.16

Close VueScan when the second scan completes.

4.17

Wait while Zooprocess converts and analyzes the background scans.

5. Aliquoting Sample

5.1

Scanning aliquots are removed from the volume-standardized, size-fractionated samples using a pipette. Three scans are performed per net sample: two scans of equal aliquot volume are done of the small size fraction (200-999 µm) and one scan is done of the large size fraction (≥ 1000 µm).

5.2

The volume of the aliquot depends on the density of the sample. There should be 700 to 1200 organisms per scan for the large fraction and 1500 to 2000 organisms per scan for the small fraction.

5.3

The aliquot volumes used are recorded in the scanning log.

6. Scanning Sample

6.1

Fill the scanning area of the Zooscan with deionized water and place the narrow frame in the scanning area.

6.2

Pour the scanning aliquot inside the frame on the Zooscan scanning area.

6.3

Use wooden probes to separate the organisms so that no organisms are touching the edges and so that no body parts from one organism are touching any other organism.

6.4

In Zooprocess, choose SCAN Sample with Zooscan (for archive, no process) process.

6.5

Open VueScan.

6.6

Press the light initialization button on the Zooscan.

6.7

Push the preview button in VueScan to preview the background image. Check the preview image to make sure that there are no lines or spots on the images.

6.8

Wait at least 30 seconds after pressing the preview button, and then push the scan button in VueScan.

6.9

Push the OK button on the Zooprocess dialog box.

6.10

Close VueScan when the scan completes.

6.11

The sample should be named in the following style: "c0810_080_051_1026_0003_l_a" where 0810 represents the year (08) and month (10) of the tow, 080 is the CalCOFI line, 051 is the CalCOFI station, 1026 is the month (10) and day (26) of the tow, 0003 is the hour (00) and minutes (03) when the tow was taken, l is the size fraction (s if it is from the small fraction), and a is the replicate (a for the first scan from a fraction, b for the second scan from a fraction).

6.12

Remove the narrow frame from the scanning area and rinse any organisms back into the scanning area using deionized water.

6.13

Slowly lift the Zooscan scanning area and pour the sample into the dish below the Zooscan.

6.14

After all samples from a tow are scanned, recombine the scanned aliquots, the large size fraction, and the small size fraction using a 202 µm plankton sock.

6.15

Return the recombined sample to the original sample jar and fill the jar with 5% sodium tetraborate buffered Formalin.

7. Image Processing

7.1

At the end of the scanning day, all scans done that day need to be processed. Processing involves normalizing the background, identifying regions of interest (ROIs), and measuring ROIs.

7.2

In Zooprocess, select "CONVERT & PROCESS Images and organisms in batch mode" process.

8. Automatic ROI classification

8.1

If it has not been done previously, a learning set will need to be created and tested by manually sorting at least 200 images from each category that you want to use in the classifier. The learning set is tested by manually sorting several hundred randomly selected images and using Tanagra and Plankton Identifier software to determine how well the learning set is able to predict the classification of the test set (see Gorsky et. al, 2010 for detailed explanations).

8.2

When all scans for a given project have been completed, copy all of the .pid files to the Unsorted_vignettes_pid subfolder of the PID_process folder.

8.3

Open Plankton Identifier software.

8.4

Select the Data Analysis option in Plankton Identifier.

8.5

In the Original Variables column, uncheck the boxes for the following variables: "X", "Y", "XM", "YM", "BX", "BY", "XStart", "YStart", and Tag.

8.6

In the Customized Variables column, uncheck the box for "ESD".

8.7

Select the appropriate learning set (large or small).

8.8

Select at up to 10 samples (make sure to select only small samples if using the small learning set and large samples if using the large learning set).

8.9

Select the Random Forest classification method (in Ohman lab, called random forest 100). If no Random Forest classifier exists, it will have to be created (see Gorsky et al., 2010 for details).

8.10

Make sure that the "Save detailed results for each sample" option box is checked.

8.11

Click the Start Analysis button.

8.12

Rename the file so that it matches the name of the sample. Save the file in the Pid_results subfolder of the PID_process folder.

8.13

Wait while Plankton Identifier classifies the ROIs.

9. Manual verification of automatic classification

9.1

Open Zooprocess software

9.2

Choose "EXTRACT vignettes in folders" ("prediction" or "vis" folder process).

9.3

In Zooprocess dialog box, leave all default settings.

9.4

Click OK in the rename subfolder dialog box.

9.5

When Zooprocess has finished extracting the ROIs, they will be in a subfolder named coded with the date and time that the ROIs were extracted in the Prediction subfolder of the PID_process folder.

9.6

Manually, check each image in the classification subfolders. If an image is misclassified, move it to the correct classification subfolder.

9.7

After all ROIs are sorted into their correct classification subfolders, choose the "LOAD identifications from sorted vignettes" process in Zooprocess.

9.8

The text files containing all of the measurements and the correct identifications are in the Sorted_vignettes subfolder of the PID_process folder. These files can be used to calculate abundances, size distributions, etc.

10. Sample Calculations

Measurements from Zooprocess are given as pixels, so it is necessary to convert the pixels to millimeters.

For linear measurements:


measurement (mm) = measurement (pixels) * 0.0106

For areal measurements:


measurement (mm2) = measurement (pixels) * 0.01062

11. Equipment/Supplies

  • Zooscan hardware
  • ImageJ software
  • Zooprocess software
  • Plankton Identifier software
  • VueScan software
  • Tanagra software
  • 202 µm and 1000 µm stacked sieves
  • 202 µm nitex mesh plankton concentrating sock
  • ring stand
  • Pint and quart jars
  • glass cleaner

12. Reagents

  • 5% sodium tetraborate-buffered Formalin in sea water
  • deionized water

13. References