MatriXense Studio
MatriXense Studio measures the matrix, the forces and the interactions of cells in 3D. Automatically, without labels, on your own computer.
Lessons: how cells move, and how they pull. Turn your sound on. Or see why it matters
Illustrative model for demonstration. Real measurements come from your bead and brightfield z-stacks.
Made for every lab that grows cells in 3D: academic research groups, imaging core facilities, and biotech and pharma teams.
Why 3D
Most cell biology still happens on flat, rigid plastic. Cells in the body live inside a soft, fibrous 3D matrix, and they behave differently there.
Human cells in 3D can also answer questions that once needed an animal. How this supports the 3Rs ›
Further reading from our founder: Sapudom & Pompe, Biomater. Sci. 2018 · Alatoom et al., Adv. Biol. 2023 · Quartey et al., Mechanobiol. Med. 2024
Cells sit at different depths, so every timepoint is a z-stack. One well, 12 h every 10 min, 50 z-slices, 2 channels = 7,200 images, about 15 GB. A 96-well plate = 691,200 images, about 1.4 TB. Manual analysis cannot keep up, and important effects go unmeasured.
Data size for 1024 × 1024 pixel, 16-bit images. Larger cameras give up to four times more.
How it works
MatriXense Studio follows your 3D cell study through four steps, one module each, and is designed to bring the results together for the same gel.
Do denser, more aligned regions carry higher strain energy?
Do cells that pull harder migrate faster, or attract more partners?
How do contacts and contraction remodel the network around them?
Step 1 · The matrix
Module · 3D FibreNet
3D FibreNet measures fibres, pores and alignment in 3D. Pore size depends on how you measure it, so it shows all seven methods and how far each one is from the truth.
Pore size, seven ways
Range of error against 3D truth over 4 simulated networks. Hover a row for values.
Whole-fibre length traced through crossings, diameter corrected for the microscope, waviness and junctions.
Every common definition side by side, including percolation and Voronoi.
In-plane alignment and out-of-plane order of the network.
3D values with 80 % intervals, inferred from simulated networks with known truth.
Step 2 · The cells
Module · Label-free 3D Tracking
Image your cells in brightfield z-stacks. The software finds every cell, places it in 3D and follows it through time, without a dye, a reporter or a transfection.
Brightfield z-stacks over time, on the microscope you already have.
Each cell's own focal plane from its phase signal, between slices.
Outline, area, volume and dry mass for every cell.
3D paths with divisions and lineages. A second opinion checks each link and asks you only where it is unsure.
No staining or transfection that can change how they move or pull.
Low-intensity brightfield instead of hours of fluorescence excitation.
Patient-derived and hard-to-label cells, from day one.
Keep fluorescence for beads, markers or a dead-cell stain.
Cells found in brightfield in validation tests: simulated cells placed into real images.
Fewer tracking errors on fast cells, in simulation.
Of the links the assistant is sure about were right, so you confirm those in one check and spend your time on the hard ones (simulation).
Then forces, interactions and community behaviour come from the same tracks, all in one app. See the forces they make
Step 3 · The forces
Module · 3D Traction Force
3D traction force microscopy from beads in the gel: how each cell deforms its matrix, in 3D and over time.
Measured example
Fibroblast in 3D collagen, 13 timepoints over 2 h, 99 % of the signal above the noise floor. Gel stiffness E = 1 kPa assumed.
Strain energy one cell puts into its matrix after 2 hours.
Local matrix densification within 25 µm of the same cell.
Bead localisation noise in x and y, so small pulls still count.
Every bead found and matched between timepoints with a calibrated 3D fit.
Each map shows which voxels rise above the measured noise, not just colour.
Joint force fitting gives each cell its own pull, even when cells touch.
Stage and gel drift are removed on coordinates, so forces stay sharp.
Step 4 · The interactions
Module · Cell Interactions
Pursuit, contact, avoidance, moving together. Detected for every pair, in every frame, without a single label.
Simulated example. In the Cell Interactions module, each interaction is also tested against randomised tracks and an interaction model fitted to your population.
Every pair is labelled same or different type. Partners joining and leaving, episode durations, and whether cells prefer their own kind.
Pursuit, attraction, avoidance, alignment and contact killing, plus the forces between cells and the matrix bridge they build.
Each index is compared with the same tracks rotated or shifted in time at random, and with an interaction model fitted to the population.
Applications
The same measurements power drug screening, mechanophenotyping and large studies across plates.
Application
Test compounds on cells inside a matrix, not on plastic. Read out contractility, migration and interactions for every well. Which drugs?
Tap a row label to see its dose response.
Application
Give every cell a mechanical fingerprint: how hard it pulls, how it moves, how it reshapes and remodels its matrix, and whom it touches. Why it matters
Each dot is one cell, placed by its forces, shape, motion, matrix and contacts.
Mechanophenotyping · why it matters
How a cell pulls, moves and reshapes its matrix is a functional readout of what it is doing. It needs no label, it reflects behaviour in a tissue-like setting, and it can reveal differences that marker staining alone misses.
Questions it helps you answer
Where it is used
Drug screening · why it matters
On stiff plastic, cells adapt to an unnatural environment, which can change how they respond to a drug. In a 3D matrix you see the effect on what matters in disease: how cells contract, move, remodel their surroundings and interact.
Questions it helps you answer
Relevant drug classes
lower higher changed Expected direction of the effect of interest.
The classic collagen gel contraction assay, measured cell by cell in 3D instead of by the shrinking gel diameter.
3D Traction ForceCAR-T, NK and T cells attacking target cells in 3D: search, contact, killing, and how long each step takes.
Cell InteractionsLink scaffold architecture to cell forces and migration, then tune stiffness, pore size and alignment on purpose.
3D FibreNet + Traction ForceCheck collagen lots and gel batches for fibre density and pore size before an experiment starts, not after it fails.
3D FibreNetFollow how cells change their forces, shape and movement when mechanical load is removed or altered.
Tracking + Traction ForceFitted interaction models and simulated twins of your network turn measurements into predictions you can test.
Cell Interactions + 3D FibreNetBatch and big data Beta
Point it at a folder of plates, set the analysis once, and let it run overnight. Every well is analysed the same way and lands in one table.
Batch analysis is in beta. It works today, and we are refining it with early users. Early users get a reduced beta price, locked for their first year.
Plate, phenotype map and batch figures are illustrative examples, not measured data.
Translational research
Put a patient’s own cells into a 3D matrix and measure how they respond, in the lab, before anyone is treated.
Test treatments ex vivo on patient-derived fibroblasts, tumour cells or immune cells, and compare who responds.
Label-free, functional readouts of how therapeutic cells move, find, contact and kill, to support potency assay development.
Find mechanical signatures linked to disease, such as activated fibroblasts in fibrosis or invasive cells in cancer.
Human 3D data to guide decisions before animal or clinical studies, in line with the move toward new approach methodologies.
Illustrative pathway. MatriXense is for research use only and not for use in diagnostic procedures.
The 3Rs
Human-cell 3D models, measured in depth, support the 3Rs first defined by Russell and Burch in 1959: Replace, Reduce and Refine animal experiments.
Ask mechanistic questions in 3D matrices with human primary or patient cells, where a 3D model can answer them instead of an animal.
Get dozens of measurements from every cell, label-free and from the same sample, and screen widely in 3D first. Only the strongest candidates go on to animal studies.
Arrive at any animal study with a clear mechanism, a dose range and defined endpoints, so it can be smaller and better targeted.
Try every module on your own images, free. We set it up with you and stay with you until your first results.
Accuracy
Every method is validated against simulated data with known truth before it ships.
Bead localisation noise in x and y, measured on real time-lapse data.
Force error per cell when neighbours are close, with joint fitting, against 47–232 % for fitting each cell alone (simulation).
Cells detected in brightfield without a stain, in validation tests with simulated cells placed into real images.
Lower z-velocity error after 3D smoothing with per-axis noise.
Cell divisions called correctly in simulation, building full lineages.
Pore-size definitions compared side by side in 3D FibreNet.
Go deeper than averages. Find what makes each cell type different, and show it.
Rhythms you cannot see by eye
Cell spectroscopy finds how each cell pulses, probes and changes shape over time.
Fibroblasts
Macrophages
Example with simulated data. Wavelet power of cell shape over time; outlined = significant above noise; faded = edge of the recording. The strip on the right is each cell type's spectral barcode.
Differences, tested
Statistics, typical cells and a figure at journal size, in the same app.
Example with simulated data. Welch t-test, Holm-corrected · ring = typical cell (closest to the median) · saved at 89 mm, 600 dpi.
Welch t-test or Mann–Whitney, against a reference or all pairs, Holm or Bonferroni. Stars and a table, checked against SciPy.
The app picks the image and the cells closest to each condition's median for your figures.
A short audit of about 30 clicks gives an honest accuracy for your trained model, and tells you when to stop labelling.
Graphs as PNG or SVG at journal size. Tables as CSV, Excel and Prism, with a dictionary for every column.
The science
Our tools start from methods published in peer-reviewed journals. We keep improving them, and every new version is checked against simulated data with known truth.
Privacy
The analysis engine runs locally and reads your images where they are. Nothing is uploaded, copied or sent anywhere.
A normal Finder or Explorer dialog. LIF, CZI, ND2, OME-TIFF, Micro-Manager, BioTek Gen5, Opera and more.
NVIDIA GPU workstations, Apple Silicon Macs, or CPU only.
Your licence is a signed key file locked to your computer. No account, no connection needed.
Our story
We believe the next breakthroughs in biology and medicine will come from seeing how cells behave in a tissue-like 3D world, not on plastic. MatriXense exists to make that world measurable: reliably, simply, and for every lab.
Collagen network topology: pore size, fibril diameter, fibril content.
Label-free 3D cell tracking in brightfield z-stacks.
3D FibreNet: 3D fibre graphs, seven pore methods, 3D values from sparse stacks. 4D MechanoInteractome: forces, lineages and community behaviour.
Deep-learning segmentation that runs on GPU, Apple Silicon or CPU, fibre tracing through junctions, and a nonlinear force solver for fibrous collagen.
Founder. Biochemist (Dr. rer. nat., Leipzig University, Germany) working in immunobioengineering and mechanobiology for more than a decade: 3D collagen matrix models, tissue–immune interactions, fibrosis, the tumour microenvironment, ML phenotyping and digital twins. MatriXense exists because the analyses he needed did not.
Licences
Every lab asks a different question. Tell us what you want to measure, and we quote a yearly licence for exactly the modules and analyses you need. No usage limits.
For institutes
Core facility & site licenceA licence server for many users, usage reports per group, and staff training.
Request a quoteTry it on your data
Pilot analysisImage quality report, full analysis and results. The full fee is credited toward a licence within 3 months.
Send a pilot datasetWe analyse it for you
Study and screening analysisTables, figures and a methods text ready for your paper. No software needed.
Get a quoteEvery licence includes all updates, new methods and improved models for as long as it runs.
Analyse as many images, wells and plates as you like. No counting, no surprise bills.
If a licence ends, projects, viewers and exports keep working. Only new analyses pause.
Services
Custom tools, analysis and advice from the people who build the software.
Every project is quoted up front after a free scoping call.
Analysis tools made for your assay and microscope, with a GUI your team can use.
Price on request
Describe your toolSend your images. Get tables, figures and a methods text ready for the paper.
Price on request
Send a pilot datasetAdvice on 3D collagen models, traction force experiments, imaging and ML phenotyping.
Price on request
Book a callFAQ
Everything you might ask before your first run.
No. MatriXense Studio runs on your own computer and reads your images where they are. Nothing is uploaded, copied or sent. For the analysis service, data moves the way your institution prefers.
LIF, CZI, ND2, OME-TIFF and TIFF series, plus folder layouts from Micro-Manager, MetaMorph, BioTek Gen5, ImageXpress, Opera/Operetta and CellVoyager. Point it at a folder; it works out time, channels and z for you.
No. Cells are found and tracked in 3D from brightfield z-stacks alone. Keep your fluorescence channels for beads, markers or a dead-cell stain.
A recent Mac with Apple Silicon or a Windows or Linux workstation. An NVIDIA GPU speeds up deep-learning steps but is not required.
Fluorescent beads in the gel plus brightfield or a cell label, as z-stacks over time. A 2–5 µm z-step gives clearly better z forces and cell positions than 10 µm. We advise on settings before you image.
No. Every yearly licence includes all updates, new methods and improved models.
Yes. Start with one module and add others whenever you need them. Your projects and results carry over, and the licence is updated for the rest of the year.
Yes. Try every module free, on your own computer and your own data, with a personal onboarding call. Or send us one dataset as a pilot analysis, fully credited toward a licence within 3 months.
One yearly licence for the modules and analyses you choose, quoted before you buy. Updates and email support are included. Add modules during the year and you pay only for the remaining months. Universities can pay by invoice or purchase order.
Projects, viewers, graphs, movies and exports keep working. Only new analyses pause until you renew.
Every method starts from peer-reviewed work and is validated against simulated data with known truth. Each run also reports image quality, noise floors and which results are reliable.
Not from confocal width alone, because the apparent width is set by the microscope’s resolution. 3D FibreNet reports a relative thickness index that can be calibrated against electron or super-resolution microscopy.
Yes. A pilot runs one dataset through quality control and full analysis, so you see the results before committing. Full studies and screens are quoted up front for your project.
Contact
Tell us about your project and how to reach you. We reply by email. Write to us directly at
hello@matrixense.com