Warning: Active development. Cross-validate results with other software before drawing conclusions. API and file formats may change without deprecation.
FLIMKit is a Python toolkit for FLIM data from FLIM microscope systems and common TCSPC / time-tag formats (PicoQuant .ptu, Becker & Hickl .sdt, ISS time-tag, Photonscore .photons). Built as a drop-in for FLIM microscope software, with two workflows:
- Reconvolution fitting: mono/bi/tri-exponential lifetime fitting with full IRF deconvolution, per-pixel and summed modes, multi-tile ROI stitching, batch processing, timelapse and z-stack analysis (a stack fitted as one FOV with a shared reference lifetime), a settable fit window with exclusion bands (drop reflection peaks mid-decay), and optional time-varying background correction (FLIMfit-style, from a measured fluorophore-free reference)
- Lifetime distribution fitting: Gaussian and Lorentzian continuous α(τ) distributions (Lakowicz §4.11.2), per-ROI and per-pixel maps with GPU acceleration
- Phasor analysis: calibrated phasor plots, interactive elliptical cursors, spatial filtering (gaussian/median/wavelet), two-component decomposition, automatic peak detection, session save/load
Four entry points: desktop GUI, guided terminal UI, CLI scripts, Python API.
Examples repo | Full documentation
Desktop GUI showing reconvolution fitting with per-pixel lifetime maps, summed decay curve, and full IRF deconvolution. The residual graph now displays beneath the tail fit for detailed fit quality assessment.
Python ≥ 3.12 required (3.14 recommended, official builds use 3.14).
git clone https://github.com/alex1075/FLIMKit.git
cd FLIMKit
python install.py # auto-detects GPU and installs the right backend
python validate_installation.py # 10 checks - all should passFor development work (PyInstaller + test dependencies):
python install.py --devOr download the compiled app from the Releases tab (no Python needed).
A pre-built image is available on Docker Hub. It runs the full desktop GUI in a browser via xpra (no installation needed on the client).
Pull and run:
docker run -d \
-p 14500:14500 \
-v /path/to/your/data:/data \
--name flimkit \
alex1075/flimkit:latestThen open http://localhost:14500 in your browser. PTU files and data should be placed in the folder you mount to /data - use the file dialog inside the app to navigate there.
TrueNAS SCALE (Custom App):
- Apps → Discover Apps → Custom App
- Paste the contents of
docker/docker-compose.yamlfrom this repo - Edit the volume paths to match your pool (e.g.
/mnt/tank/microscopy:/data) - Deploy - TrueNAS will pull the image automatically
Build from source (required if you want to push your own changes):
python docker/build_docker.py # builds linux/amd64, pushes to Docker HubRequires Docker Desktop with buildx. On Apple Silicon, buildx cross-compiles for linux/amd64 automatically.
python main.pyFive tabs: Single FOV Fit, Tile Stitch / Fit, Batch ROI Fit, Machine IRF Builder, Phasor Analysis. The right panel shows an FOV preview (intensity image + summed decay) and switches to the interactive phasor view when that tab is active. The Single FOV Fit tab has an Analysis toggle next to the input file: switch it to Z-stack to pick a folder of region_zX.ptu slices (one PTU per z-slice, e.g. a Leica .sptw workspace) and fit the whole stack as one FOV with a shared reference lifetime. The FOV Preview shows the intensity when you pick the folder and the fitted FLIM after fitting, with a z-slider (shown only for a z-stack) to scroll through depth. In a project folder, a z-stack appears as one collapsed entry (a Z tag) in the browser and reloads its fitted stack when selected.
python main.py --clipython fit_cli.py --ptu data.ptu --machine-irf machine_irf_default.npy --nexp 2
python phasor_cli.py --ptu data.ptu --irf irf.xlsxGenerates FLIM data with a known ground truth (sample PTU + matching IRF PTU + a JSON of the parameters used) for cross-software validation. Optional reflection peaks, pile-up and background let you reproduce specific artefacts; --sdt also writes Becker & Hickl versions.
python synth_cli.py --out ./validation --tau 3.0,0.8 --amps 0.7,0.3 --photons 1e5from flimkit.phasor_launcher import launch_phasor
state = launch_phasor('data.ptu', irf_path='irf.xlsx')
# With spatial phasor filtering
state = launch_phasor('data.ptu', irf_path='irf.xlsx',
phasor_filter='gaussian', filter_kwargs={'sigma': 1.5})FLIMKit auto-detects the file type and routes every format through one loader (FLIMFile in flimkit.formats), so the fitting, phasor, stitching, and ROI workflows are identical whatever the instrument wrote.
| Format | Extension | Loads as | Reader | Validated against real files |
|---|---|---|---|---|
| PicoQuant PTU | .ptu |
Fitting + phasor | ptufile |
Yes (32 files) |
| Becker & Hickl SDT | .sdt |
Fitting + phasor | sdtfile |
Yes (bit-identical) |
| Photonscore LINCam | .photons |
Fitting + phasor | photonsfile |
Yes (bit-exact vs SDK) |
| PicoQuant BIN | .bin |
Fitting + phasor | ptufile |
Upstream |
| PicoQuant PHU | .phu |
Fitting (no image) | ptufile |
Upstream |
| SimFCS B&H | .b&h |
Fitting + phasor | lfdfiles |
Upstream (no time axis in file) |
| SimFCS BHZ | .bhz |
Fitting + phasor | lfdfiles |
Upstream (no time axis in file) |
| ImSpector FLIM TIFF | .tif, .tiff (sniffed) |
Fitting + phasor | tifffile |
Upstream |
| ISS Vista TDFLIM | .iss-tdflim, .tdflim |
Fitting + phasor | lfdfiles |
Upstream |
| FLIM LABS imaging | .json (sniffed) |
Fitting + phasor | phasorpy |
Upstream |
| ISS time-tag | .tagtime, .tagchannel, .tagdecay |
Fitting + phasor | FLIMKit (from ISS spec) | No |
| ISS FD-FLIM | .ifli |
Phasor only | lfdfiles |
Upstream |
| SimFCS referenced | .ref, .r64 |
Phasor only | lfdfiles |
Upstream (no frequency in file) |
| PhasorPy OME-TIFF | .ome.tif (sniffed) |
Phasor only | tifffile |
Upstream |
| FLIM LABS phasor | .json (sniffed) |
Phasor only | phasorpy |
Upstream |
| ISS intensity image | .ifi |
Intensity only | FLIMKit (from ISS spec) | No |
"Upstream" means decoding is delegated to a maintained third-party reader that is tested against real files by its own author; FLIMKit has not independently re-validated it. "Sniffed" means the extension is ambiguous, so the file is identified by content, not by its name.
Formats whose files carry no time axis (.b&h, .bhz) or no modulation frequency (.ref, .r64) will prompt for the missing value, since fits and the universal circle cannot be computed without it.
PicoQuant .ptu (T3 mode):
- PicoHarp T3 (Leica FALCON / STELLARIS)
- HydraHarp v1/v2 T3
- TimeHarp 260 N / P T3
- MultiHarp / generic T3
Decoding and image reconstruction use Christoph Gohlke's ptufile. FLIMKit's original T3 decoder, checked against ptufile on 32 real files (Leica, PicoQuant, Chroma, Zeiss; image and point-mode), is kept as a reference in the separate flim-native-decoders repository (see flimkit/formats/PTU/NOTICE.md).
Becker & Hickl .sdt (SPC TCSPC): the histogram / image .sdt files SPCM saves (per-pixel decays already binned). Read with Christoph Gohlke's sdtfile; FLIMKit's own decoder, written from Becker & Hickl's official SPCM documentation and checked bit-for-bit against sdtfile on real files, is kept as a reference in flim-native-decoders. Thanks to Becker & Hickl for supporting this (see flimkit/formats/BH/NOTICE.md).
Photonscore .photons (LINCam, D7 container): the position-sensitive photon-counting files from Photonscore's LINCam systems. Read with photonsfile, a pure-Python reader for the D7 container (paged protobuf, seed plus delta-coded photon streams) spun out of FLIMKit, with no native dependency and validated bit-exact against the Photonscore SDK. Each photon carries an (x, y) position and a TCSPC micro-time, so the image is formed by binning the positions into a pixel grid and the decay by histogramming the micro-time; dt calibration comes from the file's TacChannel attribute. Thanks to Photonscore for supporting this and for open-sourcing the D7 format (see flimkit/formats/PS/NOTICE.md).
ISS .TAGTIME + .TAGCHANNEL + .TAGDECAY (FastFLIM / Vista, time-domain): the three time-tag files are read together (point the loader at any one of them or their shared basename). ISS records explicit frame, line, and pixel markers, so the per-pixel reconstruction is exact.
ISS .ifi (intensity image): ISS's plain intensity-image export (VISTAIMAGE header, float pixels per channel and frame). No lifetime data, so it loads as an intensity image only (no fitting or phasor).
ISS .ifli (frequency-domain / FD-FLIM): ISS's FD-FLIM lifetime-image export (VistaFLImage header). This is already phasor data (per-pixel phase and modulation at each modulation frequency), so it loads straight into phasor analysis with fitting disabled - there is no decay to fit. Decoding is delegated to Christoph Gohlke's lfdfiles (VistaIfli), which handles the version-dependent header and the multi-position / spectral axes; FLIMKit applies the file's reference-sample calibration on top. ISS's time-domain .iss-tdflim / .tdflim archives are read through the same library.
The ISS time-tag and
.ifireaders are experimental and need testing. The.TAGTIME/.TAGCHANNEL/.TAGDECAYtriplet and.ifireaders were written from ISS's format specifications and have not been validated against real ISS acquisitions (byte order and the marker conventions are assumptions). Treat their results as unverified and cross-check them. If you have ISS data, trying it and reporting back is very welcome. The.ifliand.tdflimpaths are delegated tolfdfilesand so inherit that library's own testing. PicoQuant.ptu, Becker & Hickl.sdtand Photonscore.photonsare validated against real files.
Imaging files are reconstructed into a per-pixel decay cube (Y, X, H) from their markers (PTU scan-line markers, B&H image blocks, ISS frame/line/pixel markers) or, for Photonscore, from each photon's (x, y) position; the intensity image is that cube summed over the time axis.
Read as a decay only (no image): image reconstruction needs scan / line markers, so FLIMKit fits the decay but cannot rebuild a FLIM image from a file that has none. This covers single-spot, point, and FCS acquisitions (for example PicoQuant TimeHarp 260P point measurements), and any file exported without imaging markers.
IRF sources:
- Measured / scatter IRF from a
.ptu - Leica IRF from the exported
.xlsx(interpolated or analytical model) - PicoQuant SymPhoTime Check file (
.pck) histogram - A built machine IRF, or an analytical Gaussian IRF
A .pck IRF must come from the same instrument and TCSPC resolution as the data being fit.
Not supported yet:
- T2-mode PTUs. FLIMKit decodes T3 mode only (one TCSPC histogram per sync period). T2 records are raw global timestamps with no per-period decay, so a T2
.ptuwill not produce a meaningful decay. - Older PicoQuant formats (
.pt3,.ht3,.pt2) and Becker & Hickl raw.spcphoton streams. - Leica
.lifand proprietary LMSCOMPRESSED blocks. Export.ptufrom LAS X instead.
Before fitting, build a machine IRF for your system once and reuse it across sessions. You need matched .ptu + .xlsx pairs, 10-20 is a good number.
In the GUI, go to Machine IRF Builder, point it at your pairs folder, and save as machine_irf_default. From source this goes to flimkit/machine_irf/; compiled app saves to ~/.flimkit/machine_irf/.
from flimkit.FLIM.irf_tools import build_machine_irf_from_folder
build_machine_irf_from_folder(
folder="/path/to/pairs",
align_anchor="peak",
reducer="median",
save=True,
output_name="machine_irf_default",
)Per-pixel fitting uses a batched matrix solver that runs on GPU when a supported backend is detected. This applies to both single-FOV and tile-ROI pipelines. The same fit_per_pixel() function is used in both.
| Backend | Hardware | Notes |
|---|---|---|
| MLX | Apple Silicon (M1/M2/M3/M4) | Detected automatically |
| PyTorch MPS | Apple Silicon | Fallback if MLX not installed |
| PyTorch CUDA | NVIDIA | pip install torch --index-url https://download.pytorch.org/whl/cu126 |
| PyTorch ROCm | AMD | pip install torch --index-url https://download.pytorch.org/whl/rocm6.2 |
python install.py detects your hardware and installs the right backend automatically. No extra flags needed at runtime.
Limitations: --free-tau-perpixel mode uses batched Adam on GPU when a backend is available (n_exp ≥ 2). fit_summed (single global fit) is always CPU - it's fast enough not to matter.
Compiled app and GPU: The compiled app bundles whatever GPU libraries are installed on the build machine. A binary built on Apple Silicon will have MLX/MPS; one built on a CUDA machine will have CUDA. If you need GPU in the compiled app, build it yourself on the target hardware. See Compiled App.
Not strictly necessary, but useful after making code changes.
python install.py --dev # install test requirements first
cd flimkit_tests
python run_tests.py # all tests
python run_tests.py -c # with coverage report
python run_tests.py integration # integration tests only| Format | Description |
|---|---|
| PNG | Intensity and lifetime map images |
| OME-TIFF | Lossless export with metadata - opens in Fiji/ImageJ |
| GeoJSON | ROI geometries and stats - imports directly into QuPath |
| CSV | Fit summaries and per-ROI statistics |
| NPZ | Session files for restoring analysis state |
Done: single FOV fitting, tile stitching, batch ROI processing, timelapse and z-stack analysis, phasor analysis, GUI, session restoration, compiled app, ROI analysis with QuPath export.
Up next: config persistence, stat histograms, auto-region detection, batch n-exp in GUI. Chemical validation and publication pending.
See ROADMAP.md for details.
Fitted lifetimes from FLIMKit will typically read slightly higher than FLIM microscope software for the same data. This is a consequence of IRF placement: FLIMKit anchors the IRF at the steepest-rise point of the leading edge, which differs from how FLIM microscope software places the IRF. The difference is systematic and reproducible across acquisitions.
Lifetime distribution fitting - Gaussian and Lorentzian α(τ) models:
Lakowicz, J.R. (2006). Principles of Fluorescence Spectroscopy (3rd ed.). Springer. §4.11.2, pp. 141-144.
Time-varying background correction - measured background decay B = V·b(t) + Z:
Görlitz, F. et al. (2017). Open Source High Content Analysis Utilizing Automated Fluorescence Lifetime Imaging Microscopy. J. Vis. Exp. (119), e55119. https://doi.org/10.3791/55119
PhasorPy: Gohlke, C. et al. Zenodo. https://doi.org/10.5281/zenodo.13862586
ptufile / sdtfile (PicoQuant .ptu and Becker & Hickl .sdt readers): Gohlke, C. https://github.com/cgohlke/ptufile, https://github.com/cgohlke/sdtfile
photonsfile (Photonscore LINCam .photons / D7 reader): Hunt, A. and A. Akram. Zenodo. https://doi.org/10.5281/zenodo.21360199
Tile stitching: Preibisch et al. (2009). Bioinformatics 25(11). https://doi.org/10.1093/bioinformatics/btp184
Cellpose-SAM: Pachitariu & Stringer (2025). bioRxiv. https://doi.org/10.1101/2025.04.28.651001
FLIMKit is designed, developed, and maintained by Alex Hunt. Anthropic's Claude AI was used as an assistant for parts of the GUI implementation, compiled app builds, code debugging, and Docker packaging; all scientific design, fitting/phasor methods, validation, and the overall architecture are the author's own work.
FLIMKit reads several instrument formats. PicoQuant .ptu and Becker & Hickl .sdt reading is delegated to Christoph Gohlke's ptufile and sdtfile libraries; thank you to Christoph Gohlke for maintaining them, and for PhasorPy, which FLIMKit uses as its phasor backbone. Photonscore .photons reading is delegated to photonsfile, which was written for FLIMKit and spun out as a standalone library so it can be used without FLIMKit. Becker & Hickl GmbH, ISS, Inc. and Photonscore GmbH supported the original readers directly: thank you to Becker & Hickl (in particular Dr. Jens Balke and Enzo Marscheck) for the SPCM file-structure documentation and sample .sdt files, to ISS (in particular Dr. Shih-Chu Liao, and to Anand Yethiraj at the University of Guelph for the introduction) for the FastFLIM / Vista format specifications, and to Photonscore GmbH for the LINCam SDK, a sample .photons file, and for open-sourcing the D7 format. FLIMKit's own .ptu and .sdt decoders (the .ptu one written from PicoQuant's published documentation, without direct input from PicoQuant) are kept as cross-checked references in the flim-native-decoders repository. Per-format provenance is in each reader's NOTICE.md (flimkit/formats/<FORMAT>/NOTICE.md).
Alex Hunt: alexander.hunt@ed.ac.uk