What this workflow helps you do
Summarize repeated image-based phase area fractions after segmentation has already been reviewed. The workflow is intended for comparing comparable percentage measurements across images or samples, not for deciding where phase boundaries should be drawn.
Use this workflow when
- Summarizing repeated phase-area percentages from several fields of view.
- Comparing the variability of one phase fraction across samples.
- Reporting mean and spread while retaining the sampling and segmentation context.
What you need before you start
- Phase-percentage values with sample, image or field labels.
- Consistent phase definitions and segmentation rules across the values being compared.
- A known percentage basis, typically area fraction from a defined ROI or image field.
Step-by-step workflow
- Import the phase-percentage table and label the sample or field represented by each dataset.
- Confirm that the same phase definition, ROI rule and percentage basis were used across measurements.
- Run statistics for the selected phase-percentage series.
- Review n, mean, median, standard deviation, CV, range and confidence interval where available.
- Check unusual values against the original segmentation and field selection.
- Export the summary and report how images, ROIs and phase thresholds were selected.
What to review in the result
- Percentages being combined represent the same phase and denominator.
- Fields of view are not selectively chosen only for a desired composition.
- Outlying values are checked against segmentation quality.
- Variation between fields is reported rather than hidden behind the mean.
Before you report or export
- Describe how many images or fields were sampled.
- Keep segmentation rules consistent when comparing samples.
- Do not interpret area fraction as volume or mass fraction without the assumptions needed for that conversion.
Limitations and interpretation
- Statistics cannot correct segmentation bias or nonrepresentative field selection.
- Image area fraction depends on thresholding, ROI and feature classification choices.
- A small number of fields may not represent a heterogeneous sample.