Measurements#

PhenoTypic writes its results as tables. This reference documents every column that appears in them, grouped by the operation that writes it: one page per measurement operation and per analyzer, listed in the sidebar by stage. Each column carries a Type badge that says how far a single value can be trusted; the Tier System page explains the badges.

Reading a measurements table#

Each row of deliverables/measurements.csv describes one detected colony. Its columns come in four blocks, always in this order:

  1. Your experimental metadata, from the --metadata file: strain, condition, replicate and the like (Metadata_Strain). See Metadata.

  2. Measurements, one block per measurement operation in pipeline order. Every column name starts with its metric family: Size_Area is in the Size family and is written by MeasureSize.

  3. Image bookkeeping: the image each colony came from (Metadata_ImageName, Metadata_BitDepth).

  4. Where the colony is: Object_Label, its bounding box (Bbox_*) and, on gridded plates, its well (Grid_*). See Shared columns.

Where tables are written#

A CLI run writes these files under its output directory:

File

One row per

Holds

deliverables/measurements.csv (and .parquet)

detected colony

every measurement column, with your metadata joined

deliverables/measurements_by_feature/<Operation>.csv

detected colony

one measurement operation’s columns, plus the metadata and position columns

deliverables/measurements_by_category/<Category>.csv

detected colony

one category’s columns, plus the metadata and position columns

deliverables/qc/qc.duckdb

one table per quality check

each check’s columns; see Quality control in the sidebar

deliverables/<Model>.csv (and .parquet)

fitted group

the pipeline’s growth-model parameters and fit metrics

deliverables/qc/curation_labels.parquet

curated colony

the category you assigned in the results viewer

Analyzers you run yourself return the same columns from analyze().

Reading an operation page#

Every operation page has the same parts:

  1. A one-line summary of what the operation measures, and a link to its parameters.

  2. The table’s facts: what one row describes, where the table is written, and how many columns the operation adds.

  3. One column table per schema, listing each column exactly as it is written. Where a header embeds the analyzed column, the page shows a placeholder (<metric>) and an example.

Operations#

Operation

Stage

What it writes

MeasureBounds

Object measurements

Extract bounding box coordinates and centroids of detected colonies.

MeasureColor

Object measurements

Measure robust colorimetric statistics for each colony.

MeasureGridLinRegStats

Object measurements

Evaluate grid alignment quality using row-wise and column-wise linear regression.

MeasureGridSpread

Object measurements

Quantify within-well colony dispersion using pairwise centroid distances.

MeasureIntensity

Object measurements

Measure grayscale intensity statistics of detected colonies.

MeasureNeighborDist

Object measurements

Measure distances to grid neighbours and to the nearest object.

MeasureOrientationZones

Object measurements

Measure absolute and radial-relative hyphal orientation by growth zone.

MeasureShape

Object measurements

Measure the form of each detected colony.

MeasureSize

Object measurements

Measure the key size magnitudes of each detected colony.

MeasureSymZones

Object measurements

Measure colony radial expansion, symmetry, and canonical radial zones.

MeasureTexture

Object measurements

Measure colony surface texture using Haralick features from gray-level co-occurrence matrices.

ExpectedVsDetectedCount

Quality control

Flag groups whose detected colony count diverges from metadata.

GridOccupancy

Quality control

Flag groups whose grid occupancy (filled cells / expected) is low.

ICC

Quality control

Flag groupby groups whose replicates have low ICC(2,1) agreement.

MaxModifiedZScore

Quality control

Flag (group, time) bins whose worst member is a robust outlier.

RelativeMAD

Quality control

Flag (group, time) bins with poor robust agreement across replicates.

ReplicateAgreement

Quality control

Flag (group, time) bins with poor agreement across replicates.

TukeyOutlierFraction

Quality control

Flag (group, time) bins with a high fraction of Tukey outliers.

LinearCapAndLagModel

Growth models

Linear-softplus growth fitter with a softplus saturation ceiling.

LinearLagModel

Growth models

Linear-with-softplus lag-phase growth fitter (no saturation).

LogGrowthModel

Growth models

Logistic-growth model fitter with regularized least-squares objective.

EdgeCorrector

Plate correction

Analyzer for detecting and correcting edge effects in arrayed colony growth.