Myelin Quantifier

napari-myelin-quantifier is a napari plugin for 2D myelinated axon analysis. It supports three related workflows:

• Detect and quantify myelin rings from binary mask layers in napari.
• Import raw label/object measurement CSV files and calculate myelin morphometric features into Excel workbooks.
• Build study-level analyses from raw CSV or calculated Excel files, compare blinded/final groups, and export combined study results.

Current version: 1.3.1

Version 1.3.1 adds study-level CSV/Excel analysis, filename-based sample IDs for blinded files such as 119.csv, resumable imports from calculated_*.xlsx, direct two- or three-way comparison plots, PCA/k-means, and user-selected output folders.

Installation

Install from PyPI:

pip install napari-myelin-quantifier

If napari is not already installed:

pip install "napari-myelin-quantifier[all]"

Install the development version from GitHub:

pip install git+https://github.com/wulinteousa2-hash/napari-myelin-quantifier.git

Plugin Panels

The plugin contributes four napari widgets:

• Myelin Rings: Quantify
• Myelin Rings: Locate by ID
• CSV Quantification
• CSV Study Analysis

Open them from the napari menu:

Plugins -> Myelin Quantifier

Mask Quantification Workflow

Use Myelin Rings: Quantify when you have a 2D binary mask layer.

Expected mask semantics:

• Myelin ring pixels are foreground, non-zero, or True.
• Background is 0 or False.
• Intact ring-shaped objects are recommended. Broken rings, solid objects, and border objects can be filtered or flagged.

If your mask is stored as RGB/RGBA or grayscale, use Quick Mask Prep in the quantification panel to create a binary Labels layer before running quantification.

Quick Mask Prep

Available quick actions:

• 1-Channel: create a single-channel image layer.
• Invert: create an inverted single-channel image layer.
• Binary (Otsu): convert the selected layer into a binary Labels layer.

The prepared binary layer is automatically selected in the quantifier mask-layer field.

Example binary mask:

Image courtesy of Bo Hu Lab, Houston Methodist Research Institute.

Ring Detection and Labeling

Each detected myelin ring is:

• assigned a unique ring_id,
• localized using centroid coordinates,
• measured for bounding box and area values,
• evaluated for topology using Euler characteristic.

Example labeled output:

Topological Validation

Euler number is used to distinguish ring-like structures from likely artifacts:

• Euler = 0: ring-like object with one hole.
• Euler != 0: solid object, fragmented object, or object with unexpected topology.

Topology illustration:

Mask CSV Output

The mask workflow can export a CSV with one row per detected ring. Columns include:

• ring_id
• centroid_x, centroid_y
• bbox_x0, bbox_y0, bbox_x1, bbox_y1
• ring_area_px
• lumen_area_px
• filled_area_px
• euler
• touches_border
• optional ring_area_um2, lumen_area_um2, and filled_area_um2 when pixel size is set

Example:

ring_id,centroid_x,centroid_y,bbox_x0,bbox_y0,bbox_x1,bbox_y1,ring_area_px,lumen_area_px,filled_area_px,euler,touches_border
1,873.8658,34.4421,857,18,890,52,380,556,936,0,False

CSV Quantification Workflow

Use CSV Quantification when you already have raw object measurement CSV files from a label/object measurement tool.

The importer supports common CSV variants, including comma-delimited files and semicolon-delimited UTF-8 files with a byte-order mark.

Required input columns, using either naming format:

• 2D Area (µm²)
• 2D Filled Area (µm²)

or:

• ring_area_um2
• filled_area_um2

or a mask export without pixel size:

• ring_area_px
• filled_area_px

When mask area columns are used, the CSV workflow copies them into the standard 2D Area (µm²) and 2D Filled Area (µm²) columns before calculating features. If micrometer-squared and pixel-area mask columns are both present, the calibrated micrometer-squared columns are preferred.

Other columns, such as Time Step, Label Index, Name (NA), and 2D Euler Number, are preserved in the processed output.

CSV UI Steps

1. Open Plugins -> Myelin Quantifier -> CSV Quantification.
2. Click Import CSV.
3. Select one CSV file or multiple CSV files.
4. Review the file table for row count, required-column detection, and processing status.
5. Click Process CSV.
6. Select a processed row in the table to view the quick summary and generate plots.

For each processed input file, the plugin writes an Excel workbook next to the original CSV:

calculated_<original_filename_without_extension>.xlsx

Example:

117.csv -> calculated_117.xlsx

When multiple files are processed together, the plugin also writes:

combined_myelin_quantification_summary.xlsx

Calculated Columns

The CSV workflow assumes:

• 2D Filled Area (µm²) is the outer filled fiber area.
• 2D Area (µm²) is the myelin ring area.
• AxonArea is the inner axon area.

Calculated columns:

MyelinatedArea = 2D Filled Area (µm²)
AxonArea = MyelinatedArea - 2D Area (µm²)
Rout µm = sqrt(MyelinatedArea / pi)
Rin µm = sqrt(AxonArea / pi)
thickness (Myelin) = Rout µm - Rin µm
G-ratio = Rin µm / Rout µm
Axon Diameter µm = Rin µm * 2

Myelin thickness is the radius difference, not the diameter difference.

Validity Checks

Each row is marked with:

• Valid Measurement
• Validity Note

A row is valid only when:

• 2D Filled Area (µm²) > 2D Area (µm²)
• AxonArea > 0
• Rout µm > 0
• Rin µm >= 0
• G-ratio is finite
• G-ratio is between 0 and 1

Invalid rows are kept in the output and annotated with the reason in Validity Note.

Axon Size Classes

The CSV workflow adds Axon Size Class:

• Thin: Axon Diameter µm < 1.0
• Medium: 1.0 <= Axon Diameter µm < 3.0
• Thick: Axon Diameter µm >= 3.0
• Invalid: invalid measurements

Excel Output

Each processed workbook contains:

• Processed_Data: original CSV columns plus calculated columns.
• Summary: one-row summary statistics for the source file.

Summary statistics include total, valid, and invalid object counts; mean, median, standard deviation, minimum, and maximum G-ratio; mean and median myelin thickness; and mean, median, minimum, and maximum axon diameter.

Plots

The CSV panel can generate:

• G-ratio histogram
• Axon diameter histogram
• Myelin thickness histogram
• G-ratio vs axon diameter scatter plot
• G-ratio vs myelin thickness scatter plot
• G-ratio by axon size class boxplot

CSV Study Analysis Workflow

Use CSV Study Analysis when you want to compare multiple samples, assign blind and final group labels, resume from calculated Excel files, make group plots, run PCA/k-means, or export one study workbook.

Supported inputs:

• raw measurement .csv files
• processed calculated_*.xlsx workbooks from this plugin
• compatible Excel workbooks containing a Processed_Data sheet with calculated columns

This means you can reopen a study later by importing the already calculated Excel files instead of processing raw CSV files again.

Study Setup

1. Open Plugins -> Myelin Quantifier -> CSV Study Analysis.
2. Click Import CSV / Excel.
3. Select raw CSV files, calculated Excel files, or a mix of both.
4. Optionally click Set Output Folder to choose where newly calculated Excel files and study exports should be saved.
5. Review or edit sample metadata in the import table:
   • Sample ID
   • Animal ID
   • Image ID
   • Blind Group
   • Final Group
6. Use the Include checkbox to control which samples enter summaries, plots, PCA, and exports.
7. Click Process Selected or Process All.

Calculated Excel files are loaded as already processed samples. Raw CSV files are processed into calculated_<name>.xlsx in the selected output folder, or next to the source CSV if no output folder is set.

By default, Sample ID is taken from the filename so existing blinded sample codes are preserved. For example, 119.csv becomes 119, S119.csv becomes S119, and calculated_119.xlsx becomes 119. Enable Prefix numeric IDs with S if you prefer numeric filenames such as 119.csv to display as S119.

Traditional Comparisons

The Compare tab is for direct two- or three-way comparisons without PCA.

You can compare by:

• Sample ID
• Blind Group
• Final Group

Choose comparison values A, B, and optionally C, then generate:

• overlay histograms
• boxplots
• G-ratio vs axon diameter scatter overlays
• G-ratio vs myelin thickness scatter overlays

This is intended for straightforward subject-vs-subject, blind-group, or final-group comparisons with overlapping plots.

Study Outputs

The Export tab can write:

• full study workbook
• combined object-level CSV
• sample-level summary CSV
• PCA/cluster results CSV

The study workbook contains object-level data, sample-level summaries, PCA/cluster results when available, and analysis metadata.

Typical Mask Workflow

1. Load your mask into napari.
2. If needed, use Quick Mask Prep.
3. Open Plugins -> Myelin Rings: Quantify.
4. Adjust filtering parameters, such as minimum ring area, minimum lumen area, and border exclusion.
5. Run quantification.
6. Optionally open Plugins -> Myelin Rings: Locate by ID to jump to a specific ring_id.
7. Export the mask measurement CSV.
8. Optionally process compatible CSV measurements with CSV Quantification.

Interface:

Notes

Area-derived diameter, radius, thickness, and G-ratio values assume approximately circular cross-sections. For highly irregular axons, inspect area-based values and validity notes before interpreting derived morphometrics.

Development

Run tests from the repository root:

PYTHONPATH=src python -m pytest

Acknowledgements

Example microscopy data used in documentation were generated by the Bo Hu Lab, Houston Methodist Research Institute.

Imaging hardware and infrastructure support were provided by the Electron Microscopy Core, directed by István Katona, Houston Methodist Research Institute.

Contributing

Contributions are welcome. Please ensure tests pass before submitting pull requests.

License

MIT License.