Closed d33bs closed 1 month ago
@jenna-tomkinson - would you have any thoughts on what a default set of rules might include currently? I referenced your work from WayScience/Durbin_lab_data#26 for the feature_thresholds
found within the tests, but I wasn't sure if these should be modified at all.
Maybe a default for this issue could look like (as qc_settings.yml
) this?
versions:
- cellprofiler: 4.2.4
thresholds:
# Set a negative threshold to identify both outlier small nuclei
# and low formfactor representing non-circular segmentations.
- Nuclei_AreaShape_Area: -1
- Nuclei_AreaShape_FormFactor: -1
# find very elongated nuclei segmentations (above mean)
- Nuclei_AreaShape_Eccentricity: 2
# find large nuclei segmentations (above mean) and low formfactor
- Nuclei_AreaShape_Area: 2
- Nuclei_AreaShape_FormFactor: -2
After discussing this more with @jenna-tomkinson, the file might look more like:
versions:
- cellprofiler: ">=4.2.4"
thresholds:
# Set a negative threshold to identify both outlier small nuclei
# and low formfactor representing non-circular segmentations.
small_and_low_formfactor_nuclei:
- Nuclei_AreaShape_Area:
- value: -1
type: zscore
- Nuclei_AreaShape_FormFactor:
- value: -1
type: zscore
# find very elongated nuclei segmentations (above mean)
elongated_nuclei:
- Nuclei_AreaShape_Eccentricity:
- value: 2
type: zscore
# find large nuclei segmentations (above mean) and low formfactor
large_nuclei:
- Nuclei_AreaShape_Area:
- value: 2
type: zscore
- Nuclei_AreaShape_FormFactor:
- value: -2
type: zscore
I really like this change! It will make it easier to update and will make sure that the correct features and thresholds are used together.
Provide the ability for people to use sets of their own QC rules or use (without much configuration) a default QC set of rules. This could come in the form of a
.yml
or.toml
file.