WIPP
Use Cases
Table of contents

Use Case 2: Stem Cell Colony Growth Computation

The task of this use case is to quantify the growth of stem cell colonies through time.

Challenges

Pluripotent stem cells exist in a privileged developmental state with the potential to become any of the cell types of the adult body. These cells grow as isolated colonies, each colony comprising tens to thousands of cells as the culture progresses. The colony growth is an indicator of the cell population’s health. Because individual colonies are much larger than the single camera Field of View (FOV), colony tracking can only be done from movies of large FOV images.

Inputs

In this case, we made a movie from a 16 × 22 grid of individual FOV (total size of one large FOV image ≈ 1 GB) with a 10% overlap between image tiles in both X in Y directions. Images were collected over time using phase contrast imaging. The actual experiment generated data for 5 days at a rate of 1 mosaic every 45 min for a total of 161 mosaics. In this example, we apply the pipeline to the first 10 large FOV frames.

Image Processing Pipeline

The pipeline to extract cell count consists of:
  • 1. creating a new collection and uploading images,
  • 2. stitching image tiles,
  • 3. intensity scaling and pyramid building,
  • 4. assembling tiles into a large FOV image,
  • 5. segmentation,
  • 6. tracking, and
  • 7. feature extraction.
WIPP Stem Cell Colony Growth Computation pipeline
WIPP Stem Cell Colony Growth Computation pipeline
The first 5 steps of the pipeline are identical to the ones in the first use case (Cell Count and Single Cell Detection). We will assume that the images are already uploaded, stitched and segmented. We focus on the steps 6 and 7 to complete the dynamic measurement (tracking). The stem cell data experiment is acquired as a sequential time-sequence dataset with 16 columns and 22 rows with a starting point in the upper left corner of the grid and a horizontal combing direction.

Image Processing - Tracking

Tracking takes labeled images in input and outputs a set of globally labeled images where each unique region of interest will have a unique label assigned to it across the time-sequence. Since we are solving the problem of colony growth, it is important to note that colonies grow and merge with each other through time. This merger leads to multiple colonies at time t become a single colony at time t+1. When a merger occurs, the WIPP tracker (Lineage Mapper) identifies the event as a fusion and reports it in the fusion matrix (one of the tracking outputs). The WIPP tracking algorithm can create a fusion lineage tree as well as the regular mitosis tree. After fusion, the fused colonies lose their identities and the newly formed colony is assigned a new label.

Select “Tracking Jobs” and click on “Create new job”. Input the desired job name, select the labeled collection and change minimum object size to 5000. Select “Enable Fusion” to allow fusion between colonies and click on “submit”.
Create Tracking job
Create Tracking job
The tracking output can be explored by clicking on "Tracking jobs" and selecting the name of the job.
Tracking job output
Tracking job output

Once the tracking is completed, the colony area is computed over the tracked collection as in the first use case. Download the output of tracking and explore the images and the associated output matrices.

The output of feature extraction is 10 CSV files with tabulated information for each colony per time frame. The user can post-process these files with any software to extract the colony growth over 10 time frames. We grouped These 10 individual files into one file to obtain one row per colony with area values sorted by time along columns and used that file to plot the results.
Cell colony growth across 10 time frames
Cell colony growth across 10 time frames
The user may also plot the relative area difference between measured and theoretical area values for the ten frames and for all colonies. The following plot filters colonies whose size is less than 50 000. The relative difference for most colonies is between ±0.2. Some portion of that difference is due to segmentation errors around colony borders (pixels might not have been included).
Relative area difference between measured and theoretical area values after 450 min.
Relative area difference between measured and theoretical area values after 450 min.

The user may explore the entire collection of 168 frames that can be found at isg.nist.gov .