Marker Tables¶

Summary¶

Marker tables are row-level measurement tables stored in an object's .data dictionary. Each value in .data is a marker-like wrapper with a .df DataFrame. These tables preserve the object, cell, ROI intensity, behaviour, attribute, or ROI-property rows that later get summarized into .summary.

Unlike the summary table, marker tables are not one row per subject. A single subject can have many object rows, cell rows, ROI intensity rows, or behaviour rows.

Where It Appears¶

Experiment.data["GFAP"].df for object or cell marker measurements.
Experiment.data["GFAP_ROI"].df for ROI intensity measurements.
Experiment.data["SCN ROI Properties"].df for ROI property tables.
Experiment.data["ROIs"].df and Experiment.data["ROIs To Draw"].df for ROI zip coordinate records.
MiniExperiment.data["table-name"].df for flat CSV imports.
DataFrameExperiment.data["name"].df when table-backed data is supplied through data={...}.

Required Fields¶

The minimum useful fields depend on the marker family, but these fields are stable in normalized PyFLASH tables:

Field	Meaning
`AnimalName`	Subject identifier, normalized from source names such as `Animal Name` when needed.
Marker metric columns	Numeric or categorical measurements. In folder-backed imports these are commonly prefixed with the marker name.

Marker tables used for ROI-aware processing normally also need:

Field	Meaning
`Region`	Concrete ROI or region instance, such as `SCN1`.
`ROI`	ROI label from the source table. It may be a concrete value such as `SCN1` or a base value such as `SCN`; summary building derives ROI bases when needed.
`Hemisphere`	Normalized hemisphere label when available.

Optional Fields¶

Common optional fields include:

Field or Pattern	Meaning
`Condition`	Added after conditions are applied.
Factor columns	Columns such as `Diagnosis` or `Sex`, also added from conditions when possible.
`ImageROI`	Image panel label aligned to the ROI drawing order.
`ROINameRaw`	Original ROI label before normalization.
`Label`	Object or ROI label from the ImageJ export.
`<marker>_Volume`, `<marker>_Surface`, `<marker>_IntDen`, `<marker>_MeanIntDen`	Per-row object morphology or intensity measurements.
`<marker>_XM`, `<marker>_YM`, `<marker>_RawXM`, `<marker>_RawYM`	Per-object coordinates.
`<marker>_DistToClosest_<other>`, `<marker>_DistToVentricle`	Distance measurements added by processing.
`<marker>_VolColoc...`, `<marker>_CPCColoc...`, `<marker>_Contains...`, `<marker>_Any...`	Colocalisation, containment, or association measurements. The exact prefixes depend on the imported pipeline output.

PyFLASH drops several generated helper columns during import when it needs to recompute them consistently. Treat row-level metric columns as data-dependent.

Example¶

| AnimalName | Condition | Region | ROI | ImageROI | GFAP_Volume | GFAP_IntDen | GFAP_CPCContains_DAPI |
|---|---|---|---|---|---:|---:|---:|
| Mouse_01 | Control | SCN1 | SCN1 | LHSCN | 10.2 | 450.0 | 0 |
| Mouse_01 | Control | SCN1 | SCN1 | LHSCN | 8.7 | 390.5 | 1 |
| Mouse_02 | AD | SCN1 | SCN1 | LHSCN | 12.1 | 520.8 | 1 |

For table-backed data:

import pandas as pd
from PyFLASH import from_dataframe

summary = pd.DataFrame({
    "Subject": ["Mouse_01", "Mouse_02"],
    "Diagnosis": ["Control", "AD"],
    "GFAP Total": [123.4, 156.7],
})

objects = pd.DataFrame({
    "Subject": ["Mouse_01", "Mouse_01", "Mouse_02"],
    "Area": [10.0, 11.0, 20.0],
})

exp = from_dataframe(
    summary,
    group_col="Diagnosis",
    subject_col="Subject",
    data={"Objects": objects},
)

Produced By¶

Experiment.importCSVs(), which reads FLASH/ImageJ CSV and ROI zip outputs.
MiniExperiment.importCSVs(), which reads a flat folder of CSV files.
from_dataframe, when passed a data={...} mapping.
Processing helpers such as closest-distance, ventricle-distance, colocalisation, and summary-building code, which can add derived columns.

Consumed By¶

Experiment.createSummary(), which aggregates row-level marker tables into the summary table.
Marker-level plot functions such as plot_histograms, plot_locations, and image/representative workflows that need marker names.
Export methods that write extended data workbooks.
ROI and distance processing helpers inside Experiment.processData().

Notes¶

.data keys are not guaranteed to exactly match source filenames. ROI intensity tables are often keyed with _ROI, and repeated batch marker names can be disambiguated.
Condition and factor columns are applied after a group list is assigned. Raw imports may not contain them yet.
Some table types are metadata inputs rather than biological marker measurements. For example, ROI property tables and ROIs coordinate tables live in .data but should not be interpreted as marker object tables.
Marker table rows can be much larger than summary tables. Use summary tables for subject-level statistics unless you need row-level distributions or spatial information.