Output Files and Results Structure

WormLib generates organized output organized by image name. Each analysis produces visualization PNGs, quantification CSVs, and a compiled PDF report.

—

Output Directory Structure

After running analysis on 230713_Lp306_L4440_11:

output/230713_Lp306_L4440_11/
├── Segmentation_Masks/
│   ├── cytosol_mask.png
│   ├── nuclei_mask.png
│   └── nuclei_outlines.png
├── Spot_Detection/
│   ├── set3_mRNA_detection_230713_Lp306_L4440_11.png
│   └── erm1_mRNA_detection_230713_Lp306_L4440_11.png
├── Heatmaps/
│   ├── set3_mRNA_heatmap_230713_Lp306_L4440_11.png
│   └── erm1_mRNA_heatmap_230713_Lp306_L4440_11.png
├── Line_Scans/
│   ├── set3_mRNA_line_scan_230713_Lp306_L4440_11.png
│   └── erm1_mRNA_line_scan_230713_Lp306_L4440_11.png
├── per_cell_mRNA_counts_230713_Lp306_L4440_11.csv
├── total_mRNA_counts_230713_Lp306_L4440_11.csv
├── Classification_Report_230713_Lp306_L4440_11.csv
└── 230713_Lp306_L4440_11_report.pdf

Visualization Outputs (PNG)

Segmentation Masks

  • cytosol_mask.png — Individual cell outlines (one per segmented cell)

  • nuclei_mask.png — Nucleus boundaries

  • nuclei_outlines.png — Nuclear outlines overlaid on brightfield

Used to verify segmentation quality before analysis.

Spot Detection

  • {channel_name}_detection_{image_name}.png — Red/blue points marking detected spots

Shows spot locations in max projection view. One file per RNA channel.

Heatmaps

  • {channel_name}_heatmap_{image_name}.png — Side-by-side figure

Left panel: Max projection image with spot density overlay Right panel: 80×80 grid heatmap showing spot abundance per cell

Quantifies spatial distribution of mRNA across the embryo.

Line Scans

  • {channel_name}_line_scan_{image_name}.png — 1D intensity profile along embryo axis

Shows RNA intensity variation along anterior-posterior axis with embryo cell labels (if classifier enabled).

Quantification Outputs (CSV)

Wide Format: ``total_mRNA_counts_{image_name}.csv``

Summary counts per image:

Image ID,set3_mRNA total molecules,erm1_mRNA total molecules
230713_Lp306_L4440_11,547,302

Useful for quick statistics across many images.

Long Format: ``per_cell_mRNA_counts_{image_name}.csv``

Per-cell spot counts:

Image ID,region_id,set3_mRNA,erm1_mRNA,label,confidence
230713_Lp306_L4440_11,1,125,89,AB,0.987
230713_Lp306_L4440_11,2,98,72,P1,0.954
230713_Lp306_L4440_11,3,156,104,ABa,0.923
230713_Lp306_L4440_11,4,168,37,EMS,0.891

Columns:

  • Image ID — Image filename

  • region_id — Cell number (1, 2, 3, …)

  • {channel_name} — Spot count per channel

  • label — Predicted cell identity (if classifier enabled)

  • confidence — Prediction confidence (0.0–1.0)

Use this for statistical analysis, correlation tests, etc.

Classification Report: ``Classification_Report_{image_name}.csv``

Cell identity predictions with feature values:

Cell_ID,label,prediction_confidence,centroid_x,centroid_y,area,eccentricity
1,AB,0.987,512.5,256.3,18500,0.45
2,P1,0.954,520.1,389.2,22100,0.52
...

Features used by the Random Forest classifier: - Centroid position (X, Y) - Cell area (pixels²) - Eccentricity (elongation) - Other morphological features

PDF Report

``{image_name}_report.pdf``

Compiled analysis report containing:

  • Segmentation visualization (masks overlay)

  • Spot detection for each channel

  • Heatmaps (left: image, right: grid)

  • Line scan plots

  • Summary statistics (total spots, per-cell distributions)

  • Classification results table (if enabled)

  • Processing parameters (voxel size, PSF, segmentation settings)

Automatically generated at the end of analysis. Suitable for sharing with collaborators.

—

Interpreting Results

High-quality segmentation

  • Clear cell boundaries in segmentation mask

  • No merged cells or fragments

  • Expected count: 2-cell or 4-cell embryos only

Robust spot detection

  • Red/blue dots in spot detection PNG match obvious fluorescent puncta

  • Not too noisy (false positives from background)

  • Not too conservative (missing real signal)

  • Adjust spot_radius_nm in config if results look off

Meaningful heatmaps

  • Grid heatmap shows clear spatial patterns (not uniform)

  • High-abundance regions correspond to visible spots in image

  • Compare between channels to identify differential localization

Cell classification accuracy

  • Check confidence column in per_cell CSV

  • Confidence > 0.9 is generally reliable

  • Lower confidence suggests ambiguous cell identity or segmentation artifact

—

Troubleshooting Output Issues

No output files generated

  • Check pipeline flags in config (e.g., spot_detection: true)

  • Verify output_directory exists and is writable

  • Check console output for error messages

Blank or noisy visualizations

  • Verify channel_indices are correct

  • Check image data (plot in Jupyter to inspect)

  • Confirm PSF values are appropriate for your microscope

CSV files missing

  • Spot detection must run before quantification CSVs are generated

  • Classifier must be enabled for label and confidence columns

  • Review pipeline flags

—

Next Steps