Case Study: Counting Cells with the Crosshairs Tool
This case study explains how to use the crosshairs tool to count cells and determine their position.
Counting many similar objects in microscopy data is a mainstay of quantitative image analysis in the biological sciences. Automated analysis is fairly routine, but accuracies can suffer when there are different types of cells in the same culture that need to be distinguished, or where cell morphology or imaging conditions are not ideal for subsequent segmentation. Human intervention is often required in such circumstances, which is easy to accomplish with Quantius.
First, we save a stack of images from different locations in a microwell plate co-culture of cells labeled with green and red nuclear fluorescent tags. Here, we are interested in retrieving the x,y coordinates and total number of only the green cells. We save the individual frames as 1.png, 2.png, . . . , 30.png into a folder using File > Save as . . . . Then, we upload the images to Quantius as a new job with the following parameters:
- Job name: Countgreencells.
- Tool: Crosshairs.
- Slider: Off.
- Instructions: "Click the center of the green objects but not the pink ones. Ignore the green objects that overlap with the edge of the image."
- Replicates: 10.
Post-processing of the raw Quantius annotations involves cluster analysis to determine the cell centers from the cloud of crowd annotations in Matlab, simultaneously eliminating rare spurious annotations.
Quantius analysis allows counting of specific objects in complex images with accuracy equivalent to trained experts. a Individual frames from an image stack are saved into a folder as a set of .png images. b Job setup parameters in the Quantius interface. c An example image with overlays of raw Quantius worker annotations and cell positions inferred from cluster analysis in Matlab. Right: a plot of annotation distance from the nearest cluster representing a nucleus center. The majority of annotations fall within around a quarter of a nuclear width from the corresponding nucleus center; and these spatial error distributions are similar for trained experts and Quantius workers.