2D Measurement of Stem Cells to Score Their Quality

Purpose: Reduce uncertainty in the quality evaluation of stem cell colonies from 2D microscopy images where the quality is assessed by expert raters to train a regression model

The visual inspection of pluripotent stem cell colonies by microscopy is widely used as a primary method to assess the quality of the preparations and degree of pluripotency. Because automated methods for culturing, expanding and differentiating stem cells are increasingly being used to reduce preparation variability, automated microscopic methods that enhance the consistency of evaluating the characteristics of colonies in situ are needed. We developed a robust approach based on Bayesian statistics for the evaluation of pluripotent stem cell colony quality based on training sets provided by two stem cell experts. Two experts rated phase contrast microscope images of human embryonic stem cell (hESC) colonies on a scale of 1 (poor) to 5 (maximum pluripotency character), but agreed with one another only 48% of the time. We developed custom image feature algorithms based on experts’ stated visual criteria for selection of colonies. These features, plus others, were then used to develop pluripotency scoring algorithms trained to reflect ratings of both experts. We treated expert ratings as inexact indicators of a continuous pluripotency score and considered the inconsistency between expert ratings in developing our models. A subset of colony ratings by experts was used for training and the remaining colony ratings were used for testing using a 50-fold cross-validation regimen. A linear model based on both experts identified each experts’ top-rated colonies as well as or better than did the ratings of the other expert, as indicated by receiver operator characteristic curve analysis. Covariance analysis indicated that both experts use features that are not included in the model. Two image features, colony perimeter and a feature based on texture, were the most important for predicting experts’ ratings. Interestingly, colony perimeter was not one of the expert-provided criteria for rating colonies.

Deep Zoom view of stitched images

This dataset contains 3 composite images of ~450 human embryonic stem cell (hESC H9) colonies acquired in phase contrast on an inverted microscope over a total area of approximately 12 cm^2. A material that allowed assessment of alignment of the phase rings was used to provide confidence of the quality of the images. Adjacent fields of view were imaged with a 10% overlap and then stitched together. Colonies within the image set were identified by automated analysis, and images of each of the colonies were presented to the experts for rating. A total of 449 pluripotent stem cell colony images were rated by two experts. The cells were maintained in culture on mouse embryonic fibroblast (MEF) feeder cells. The url points to a website where the user can zoom into the composite images to view individual colonies,and see image analysis operations (layers of results) that demonstrate image correction and segmentation to produce the colonies images that were provided to experts.

Data analysis sequence

  1. Step 1: Phase contrast acquisition and stitching
  2. zip file download of 3 raw intensity images of wells after stitching (2.5GB)
  3. Step 2: Phase contrast meniscus removal
  4. zip file download of 3 raw intensity images of wells with meniscus removed (1.4GB)
  5. Step 3: Colony segmentation and mask creation
  6. zip file download of 3 binary colony mask images of wells (1.7MB)
  7. Step 4: Image feature extraction from colony masks
  8. Step 5: Scores of phase contrast images of masked colonies by domain experts
  9. Step 6: Statistical analyses of colony features and experts' colony rankings

The source code for processing phase contrast images can be found here.

This page is currently under construction.