bleachdt 4.1.1

BleachDt - A comprehensive toolkit for analyzing FRAP experiments in phase-separated droplets

BleachDt

A comprehensive Python toolkit for analyzing FRAP experiments in phase-separated droplets

The name "BleachDt" (pronounced "bleached") is a nod to the optional PDE-based diffusion coefficient (D) fitting.

Installation • Quick Start • GUI Guide • Command Line • Parameters • Output Files

---

Features

• Fiji-matched Stabilization — Registers every frame to the pre-bleach frame on the whole-image gradient magnitude, exactly like Fiji's Image Stabilizer plugin. Verified against the ImageJ plugin's own transformation log: shifts agree to ~0.03 px. The gradient-magnitude domain makes this robust to the bleach event and photobleaching (which are not motion). A single subpixel bilinear shift is applied to all channels (estimated from the analysis channel only), so a fixed ROI tracks the true droplet area across all frames. A stabilization_shifts_*.csv log is written for QC. (Optional --exclude-bleach-from-stabilization masks FRAP droplets for very sparse-droplet movies; off by default.)
• Automated Droplet Detection — Identifies and crops individual droplets from multi-droplet image stacks, recording each droplet's X/Y location and geometry.
• PDE-Based Diffusion Fitting — Fits diffusion coefficients using partial differential equation model to compensate for bimodal recovery that single exponential fitting obscures.
• Dual Normalization — Computes BOTH the standard pre-bleach-frame normalization and the whole-droplet fluorescence per frame normalization. Raw integrated densities (RawIntDen) and both normalized curves are always saved; the GUI selects which is reported as the headline recovery curve.
• Dual Model Support — Choose between "Separate" (independent PDE + exponential) or "Combined" (constrained) fitting models.
• Consolidated Outputs — One per-movie droplet_metrics.csv merges FRAP, NoBleach, and droplet location/geometry.
• Interactive GUI — User-friendly interface with real-time, re-runnable scatter-plot filtering.
• Batch Processing — Parallel processing across multiple conditions and experiments.
• Automatic Configuration Saving — Every run saves its parameters and log output for reproducibility.

---

Installation

From PyPI (Recommended)

# Install core package (command-line tools only)
pip install bleachdt

# Install with GUI support (recommended)
pip install bleachdt[gui]

# Alternative GUI backends
pip install bleachdt[gui-pyqt6]  # PyQt6 instead of PySide6
pip install bleachdt[gui-pyqt5]  # PyQt5 instead of PySide6

Additional Microscopy Format Support

# Zeiss CZI files
pip install bleachdt[zeiss]

# Leica LIF files
pip install bleachdt[leica]

# Nikon ND2 files (included by default, but explicit)
pip install bleachdt[nikon]

# All formats (Zeiss, Leica, Nikon)
pip install bleachdt[all-formats]

# Everything (GUI + all formats)
pip install bleachdt[all]

From Source

# Clone the repository
git clone https://github.com/andrewbazley/BleachDt.git
cd BleachDt

# Install in development mode with all features
pip install -e .[all]

Requirements

• Python: 3.7 or higher
• Platforms: Windows, macOS, Linux
• Core Dependencies: numpy, pandas, matplotlib, scipy, scikit-image, tifffile, numba, opencv-python
• GUI Dependencies: pyqtgraph + PySide6 (or PyQt6/PyQt5)

Supported Image Formats

Format	Extension	Package Required
TIFF	.tif, .tiff	(included)
OME-TIFF	.ome.tif, .ome.tiff	aicsimageio
Nikon ND2	.nd2	nd2
Zeiss CZI	.czi	aicsimageio[czi]
Leica LIF	.lif	aicsimageio[lif]
Zeiss LSM	.lsm	aicsimageio
Olympus	.oib, .oif	aicsimageio
DeltaVision	.dv, .r3d	aicsimageio

---

Quick Start

Launch the GUI

# After pip install
bleachdt-gui

# Or run directly from source
python -m bleachdt.gui

Basic Workflow

1. Set Input Directory — Point to a folder containing raw movie files
2. Configure Parameters — Adjust imaging settings (pixel size, frame interval) and fitting parameters
3. Run Analysis — Click "Run Full Analysis" to process all movies
4. Filter Results — Use the Filtering tab to interactively exclude outliers
5. Export — Filtered data and configurations are automatically saved

Supported Input Formats

Format	Extension	Notes
Nikon ND2	.nd2	Full support with dimension auto-detection
TIFF/TIF	.tif, .tiff	Raw movies or pre-cropped droplet stacks
Zeiss CZI	.czi	Requires pip install bleachdt[zeiss]
Leica LIF	.lif	Requires pip install bleachdt[leica]
OME-TIFF	.ome.tif	Via aicsimageio

The pipeline automatically detects the input type and processes accordingly:

• Raw movies (ND2, CZI, etc.) → Stabilize → Crop droplets → Analyze
• Pre-cropped TIFs → Analyze directly
• Existing CSVs → Combine/filter results

---

Pipeline Components

All functionality lives in a single bleachdt package. The core modules are:

Module	Console script	Purpose
bleachdt.gui	bleachdt-gui	Main GUI application (orchestrates the full pipeline)
bleachdt.droplet_crop	bleachdt-crop	Detects, crops, and stabilizes droplets from raw movies
bleachdt.crop_pde	bleachdt-analysis	Core analysis engine for PDE fitting and curve extraction
bleachdt.combine	bleachdt-combine	Combines and filters FRAP curves across experiments
bleachdt.condition_utils	—	Condition / replicate name extraction helpers
bleachdt.r2_scatter	bleachdt-r2	R² summaries for FRAP scatter relationships
bleachdt.pre_mean_vs_radius	bleachdt-radius	Pre-bleach mean intensity vs droplet radius analysis

---

GUI Guide

The GUI provides two tabs: Analysis and Filtering.

Analysis Tab

The Analysis tab contains collapsible sections for organized parameter control:

Core Imaging Parameters

Parameter	Default	Description
Pixel Size	0.11 µm	Physical size of each pixel
Frame Interval	10.0 s	Time between consecutive frames

Image Stabilization

Applied as first step to the whole movie before droplet cropping. The stabilized full-movie TIF is saved in each movie's results folder as stabilized_[moviename].tif (e.g. Analysis Results/movie_stem_results/stabilized_movie_stem.tif).

Click to expand parameters

Parameter	Default	Description
Enable Stabilization	✓	Apply drift correction to entire movie before cropping
Algorithm	Lucas-Kanade	Lucas-Kanade (ImageJ-compatible) or ECC (OpenCV)
Pyramid Levels	4	Multi-scale pyramid levels for coarse-to-fine alignment
Max Iterations	1000	Maximum iterations per frame
Error Tolerance	1e-8	Convergence threshold (lower = stricter)
Template Update	0.90	Template update coefficient (0=fixed, 1=full update)
Pre-smooth Sigma	0.5	Gaussian smoothing before alignment (0 = none)
Max Shift (px)	25	Maximum allowed shift in pixels
Min Correlation	0.8	Minimum correlation to accept alignment (ECC)
Refine Iterations	20	Refinement passes at full resolution (ECC)

Droplet Cropping

Uses: bleachdt.droplet_crop (bleachdt-crop)

Click to expand parameters

Parameter	Default	Description
Save Droplet TIF Stacks		Save individual droplet TIF files to droplet_stacks/
Clear Borders	✓	Exclude droplets touching image edges
Min Area	500 px	Minimum droplet area to detect
Bleach Threshold	1.2	Ratio threshold to classify FRAP vs NoBleach
Crop Multiplier	2.5	Crop box size relative to droplet diameter
Min NoBleach Diameter	5.0 µm	Minimum size for control droplets
Min NoBleach PC	10.0	Minimum partition coefficient for controls
N Jobs	8	Parallel workers for cropping

PDE Fitting

Uses: bleachdt.crop_pde (bleachdt-analysis)

Click to expand parameters

Parameter	Default	Description
D Min / D Max	0.001–2.0 µm²/s	Diffusion coefficient search range
D Grid Size	25	Coarse grid points for D search
D Refine Grid	100	Fine grid points for refinement
Min Droplet Size	100 px	Minimum droplet area for analysis
Ratio Percentile	85%	Threshold for bleach ROI detection
Max Frames	0 (all)	Limit frames to process (0 = all)
Downsample Factor	1	Spatial downsampling (1 = none)
PDE Frames	1–7	Frame window for PDE fitting
Exp Frames	1–30	Frame window for exponential fitting
Enable PDE Fitting	✓	Enable PDE diffusion fitting (uncheck for exponential-only analysis)
No Photobleach Correct	☐	Skip photobleach correction
No Exp Plateau Fit	☐	Skip exponential plateau fitting
N Jobs	8	Parallel workers for processing movie folders

Detection Parameters

Uses: bleachdt.combine (bleachdt-combine)

These parameters control initial curve quality detection (not to be confused with user-defined filtering in the Filtering tab).

Click to expand parameters

Parameter	Default	Description
Max Drop	0.1	Maximum intensity drop between frames to accept a curve
Max Jump	0.1	Maximum intensity jump between frames to accept a curve
Ignore First N	4	Frames to skip in drop/jump check
Min Value	0.1	Minimum allowed intensity
Max Value	1.2	Maximum allowed intensity
Skip Detection	☐	Include all curves without detection thresholds

Output Options

• Generate Combined Output Files — Create combined CSV/XLSX files across all movies

Run Buttons

Button	Action
Crop Only	Run only droplet detection and cropping
Run Full Analysis	Complete pipeline (crop → fit → combine)
Apply Filtering Settings	Re-run analysis using Filtering tab selections (enabled after filtering changes)

Configuration Management

• Save Configuration File / Load Configuration File — Manually save/restore all parameters
• Auto-save run config ✓ — Automatically saves parameters + log output after each run

Filtering Tab

Interactive scatter plot filtering for quality control:

1. Load Data — Import FRAP_condition_summary.csv or NoBleach_summary.csv
2. Create Plot — Select X/Y parameters (e.g., D_um2_per_s vs cost)
3. Add Filters — Use horizontal/vertical lines or box selection
4. Save Filters — Apply filters to subsequent analyses

After running a full analysis, the filtering tab automatically loads summary CSVs from the analysis directory and displays the source path.

Visualization Tab

Preview individual FRAP curves from any movie in your dataset:

1. Browse Directory — Select folder containing analyzed data
2. Select Movie — Choose a movie from the dropdown
3. View Curves — See all FRAP curves plotted together
4. Apply Filters — Optionally show only curves that pass filtering parameters

This is useful for quality control to verify filtering parameters are working as expected.

---

Command Line

After installing via pip, the following commands are available:

Launch GUI

bleachdt-gui

Full Analysis

bleachdt-analysis /path/to/movies \
    --pixel-size 0.11 \
    --frame-interval 10.0 \
    --D-min 0.001 --D-max 2.0 \
    --n-jobs-conditions 8

Droplet Cropping Only

# Stabilization is ON by default using Lucas-Kanade algorithm
bleachdt-crop /path/to/movies \
    --min-area 500 \
    --min-nobleach-diameter 5.0 \
    --min-nobleach-pc 10.0

# To disable stabilization:
bleachdt-crop /path/to/movies --no-stabilize

# To use ECC algorithm instead of Lucas-Kanade:
bleachdt-crop /path/to/movies --stabilize-algorithm ecc

Combine and Filter Curves

bleachdt-combine /path/to/results \
    --max-drop 0.05 \
    --max-jump 0.05 \
    --min-val 0.2 \
    --output-prefix filtered_

Running from source (without pip install)

python -m bleachdt.gui                              # GUI
python -m bleachdt.crop_pde /path/to/movies ...     # Full analysis
python -m bleachdt.droplet_crop /path/to/movies ... # Cropping
python -m bleachdt.combine /path/to/results ...     # Combine curves

HPC / Cluster Usage

BleachDt is fully compatible with HPC environments. All scripts support parallel processing and can run entirely from the command line without a GUI.

# Run on 24 cores
bleachdt-analysis /path/to/movies --n-jobs-conditions 24

# Droplet cropping on 24 cores  
bleachdt-crop /path/to/movies --n-jobs 24

# Combine curves on 24 cores
bleachdt-combine /path/to/results --n-jobs 24

Key features for HPC:

• All processing uses Python's ProcessPoolExecutor for parallel execution
• No GUI dependencies required for command-line operation
• Progress output to stdout for log monitoring
• Configurable worker count (up to 64 by default)

---

Parameters

Primary Measurements

Parameter	Units	Description
D_um2_per_s	µm²/s	Diffusion coefficient from PDE fitting
k_single_exp	1/s	Exponential recovery rate constant
half_time_s	s	Recovery half-time, t½ = ln(2)/k
mobile_fraction	—	True mobile fraction = (YM−Y0)/(1−Y0) computed on the whole-droplet-normalized curve, so it accounts for redistribution relative to the finite droplet pool (not just bleach depth).
immobile_fraction	—	1 − mobile_fraction
recovery_amplitude	—	Raw recovered signal YM−Y0 (reported separately from the true mobile fraction)
mobile_fraction_basis	—	Which curve the mobile fraction was derived from (whole_droplet or, as fallback, fit_ROI)

Droplet Properties

Parameter	Units	Description
pre_mean	—	Normalized pre-bleach intensity
K_pre	—	Partition coefficient (dense/dilute)
bleach_fraction	—	Fraction of droplet area bleached
droplet_radius_um	µm	Effective droplet radius

Fit Quality

Parameter	Description
cost	Weighted sum of squared residuals (lower = better)
success	Boolean indicating fit convergence
n_frames_fit	Number of frames used in fitting

NoBleach Control Parameters

Parameter	Description
K0, K_end	Initial and final partition coefficients
deltaK_rel	Relative change in partition coefficient
dK_dt	Rate of partition coefficient change (1/s)
R0, R_end	Initial and final droplet radii (µm)
dR_dt	Droplet growth rate (µm/s)

---

Normalization & Raw Intensities

BleachDt computes two FRAP normalizations for every bleach spot, both in true integrated-density (RawIntDen) space with explicit per-ROI background subtraction (matching the Fiji total-fluorescence macro):

• Pre-bleach frame (FRAP_norm_prebleach): the classic FRAP double normalization. The bleach-spot background-subtracted integrated density is corrected for acquisition photobleaching using in-field NoBleach reference droplets (their mean intensity normalized to the first frame) — never the bleached droplet itself — then normalized to its own pre-bleach value. If no NoBleach reference is available the curve is left uncorrected and flagged (pb_source). The uncorrected ratio and per-frame photobleach_factor are also saved for transparency.

Background is subtracted per frame from the non-droplet region (the conventional order: local background removed at each time point before normalization).

• Whole-droplet ratio (FRAP_norm_wholedroplet): (bleach / whole-droplet) background-subtracted, normalized to the pre-bleach ratio. Accounts for the droplet's own total-fluorescence dynamics (e.g. condensate growth / influx).

Both curves and the raw RawIntDen values (bleach + whole droplet) are always written to each movie's consolidated FRAP_droplet_timeseries.csv, so you can re-derive or re-filter anything later. In the GUI, Image Stabilization → FRAP normalization selects which curve is reported as the headline recovery curve (prebleach, wholedroplet, or both); on the command line use --normalization-method. The PDE diffusion fit is normalization-independent (it fits the spatial relaxation), so this choice does not change D.

Drift is applied at subpixel precision by default (smooth registration; intensities are resampled once via bilinear interpolation). If you would rather preserve exact pixel values at the cost of up to ±0.5 px residual drift, enable Whole-pixel apply in the GUI or pass --whole-pixel-stabilization to bleachdt-crop.

Output Files

Per-Movie Outputs (Analysis Results/[movie]_results/)

File	Description
stabilized_[movie].tif	Stabilized whole-movie stack (when stabilization is on). Subpixel bilinear drift correction by default (smooth, faithful intensities; --whole-pixel-stabilization to preserve exact values instead).
stabilization_shifts_[movie].csv	Per-frame drift log: dy, dx, shift magnitude, residual RMS, method, application mode — for QC / comparison against Fiji.
droplet_metrics.csv	The single consolidated metrics table. Merges per-droplet FRAP fit parameters (incl. true mobile_fraction, recovery_amplitude, half_time_s), NoBleach control metrics, droplet X/Y location + geometry, and the scatter R²/p-values — one droplet-keyed row each. (droplet_locations.csv is folded in and removed.)
FRAP_droplet_timeseries.csv	All FRAP droplets' per-frame curves in one long CSV (droplet column). Always includes raw bleach/droplet RawIntDen, BG-subtracted integrated density, the photobleach_factor, and both normalized curves.
NoBleach_droplet_timeseries.csv	All NoBleach control droplets' per-frame K(t)/R(t) in one long CSV.
Pre-bleach_stacked_all.csv / Whole-droplet_stacked_all.csv	All FRAP curves (unfiltered) per normalization method, transposed: one row per bleach point, one column per frame.
Pre-bleach_averaged_all.csv / Whole-droplet_averaged_all.csv	Mean ± SEM recovery curve across bleach points, per normalization method.
FRAP_condition_summary.csv / NoBleach_summary.csv	Per-droplet fit table / control table (retained for the Filtering tab; their data is also in droplet_metrics.csv).
droplet_stacks/	(Optional) Cropped droplet TIF stacks; only written if "Save droplet TIF stacks" is checked.

Optional diagnostic plots (toggled by GUI checkboxes / CLI flags, applied per-movie and combined): FRAP_frap_fit_separate.tif (off), FRAP_K_exp_scatters_combined.tif (off), FRAP_noise_pixels.tif (off), NoBleach_K_curves.tif (on). The per-droplet summary/ folder is no longer created.

Combined Outputs (Analysis Results/)

File	Description
droplet_metrics.csv	Combined consolidated metrics table across all movies (adds a movie column).
Pre-bleach_stacked_all.csv / Whole-droplet_stacked_all.csv	All FRAP curves across all movies (unfiltered), per normalization method, transposed (rows = bleach points, columns = frames).
Pre-bleach_averaged_all.csv / Whole-droplet_averaged_all.csv	Mean ± SEM curve across all bleach points, per normalization method.
FRAP_condition_summary.csv	Combined per-droplet fit summary (retained for the Filtering tab).
stacked_all_Filtered.xlsx	Filtered curves (from Filtering tab).

Eliminated: the combined all_frap_curves.csv, stacked_all.csv, averaged_all.csv, and the top-level FRAP_curves.tif overlay — superseded by the transposed per-normalization stacked + averaged CSVs above.

Run Logs

File	Description
run_config_*_YYYYMMDD_HHMMSS.json	Complete parameter record + log output

---

Tips

1. Start broad, then filter — Run analysis first, then use scatter plots to identify and exclude outliers
2. Check fit quality — Plot D_um2_per_s vs cost to identify failed fits
3. Organize by movie — Each movie folder generates its own _results output folder
4. Save configurations — Use "Save Configuration File" before long runs for reproducibility
5. Monitor progress — The log output shows real-time processing status

---

Troubleshooting

Issue	Solution
Analysis failed (return code)	Open the run config JSON in the input folder; check the end of log_output for the actual error. Common causes: Droplet_Crop exception (e.g. image format not T,Y,X), or Crop_PDE "No TIF files found" (no droplets detected).
"Droplet cropping produced no droplet TIF files"	No droplets were detected in any movie. Adjust Bleach threshold, Min area, or check that stacks are (time, Y, X) and contain visible droplets.
"No CSV files found"	Ensure all_frap_curves.csv exists in result folders
"PyQtGraph not available"	Install with pip install pyqtgraph
Slow performance	Reduce D grid size or increase downsampling
Missing movies in output	Check that TIF files exist in movie folders or droplet_stacks/

---

Citation

If you use BleachDt in your research, please cite:

Bazley, A. (2026). BleachDt: A Python toolkit for FRAP diffusion analysis. GitHub repository: https://github.com/andrewbazley/BleachDt

---

License

MIT License - see LICENSE for details.

---

Contact

Andrew Bazley [e-mail]