Every discussion of MS imaging resolution arrives at the same conclusion: you can chase resolution or sensitivity, rarely both. That is true, and it is not very actionable. What helps is separating the three distinct costs that finer pixels impose, understanding which one binds in your case, and doing the arithmetic on the third — which is the only one that can be calculated exactly in advance.
Key Takeaways
|
Why Resolution and Sensitivity Compete
The primary reason is simple and worth stating without hedging: sampling a smaller area collects less material. Halving the pixel dimension quarters the sampled area and therefore roughly quarters the number of molecules of any given species available to ionise at that position. Nothing about the instrument changes that; it is a property of the sample.
What follows is that low-abundance species drop below detection first. A high-resolution image of an abundant lipid and a high-resolution image of a trace metabolite are not equally difficult experiments, even on identical hardware, and the difference is not a matter of optimisation.
Working in analytical science?
Register for a FREE Separation Science account to subscribe to the Separation Science Newsletter.
But three distinct consequences follow from finer pixels, and treating them as one thing is why optimisation advice in this area is so often unhelpful.
Trade-off | Cause | What You Lose | Available Mitigation |
Resolution against sensitivity | Fewer molecules in a smaller sampled volume | Signal for any given species, low abundance first | More shots or longer dwell per pixel, at a cost in time; post-ionisation; a more sensitive analyser |
Resolution against molecular coverage | The species still detected are increasingly the abundant and readily ionised ones | Breadth of the observable molecular range, even where signal remains | Chemistry: matrix or solvent choice, ionisation mode, derivatisation |
Resolution against acquisition time | Pixel count rises with the square of resolution | Throughput, and in the limit feasibility | Faster analyser, smaller imaged area, or accept coarser pixels |
Table 1. Three distinct trade-offs, usually discussed as one. The mitigations differ, which is the practical reason for separating them.
The second row is the one most often missed, and it is genuinely different from the first. Sensitivity is about whether a species produces a detectable signal; coverage is about how many different species remain in the detectable set. You can improve sensitivity for a target analyte by spending more time per pixel, and that does nothing to broaden coverage. Improving coverage is a chemical problem, addressed at sample preparation, which is why the deposition decision in MALDI Imaging Mass Spectrometry: How It Works turns out to be a coverage decision as much as a resolution one.
Pixel Size and Ion Yield
The relationship between pixel size and detectable signal is not only about area, and it is worth understanding the additional factors because two of them are addressable.
- Sampled volume. The dominant term. Area falls with the square of pixel dimension, and in ablative techniques depth matters too.
- Ionisation efficiency. Only a small fraction of desorbed material ionises. That fraction is chemistry-dependent and is where matrix, solvent, and polarity choices act.
- Physical sampling limits. In MALDI, matrix crystal size sets a floor: sampling cannot be finer than the crystals present. Published work is explicit that crystal size must be kept below the intended pixel size.
- Transmission and detection. Ion optics and detector efficiency set how much of what is produced is actually measured.
Two things in that list are worth acting on before accepting a sensitivity limit. Crystal size is a preparation variable, and the published comparison of matrix deposition methods shows it is decisive at fine resolution rather than marginal. And ionisation efficiency is chemically tunable, which is the lever most often left unpulled when an experiment is described as sensitivity-limited.
Resolution Is a Property of the Analyte, Not Just the Instrument This is the most useful single correction to how resolution is usually quoted. The same technique on the same instrument achieves very different resolutions depending on what you are looking for. Nano-DESI has reached around 10 microns for abundant lipids and metabolites, and around 200 microns for denatured proteins. Work demonstrating proteoform mapping down to 7 microns states plainly that a substantial decrease in protein signals is observed in high-spatial-resolution experiments, which makes those experiments challenging. So a resolution figure quoted without an analyte class is close to meaningless. When reading a specification or a paper, the question is always: at what resolution, for which species, at what abundance? The technique-specific figures are collected in Mass Spectrometry Imaging: Principles, Techniques, and Applications. |
Why Is the Time Budget Quadratic?
Because an image is two-dimensional. Halving the pixel dimension doubles the number of positions along each axis, so the total number of positions rises fourfold. Work on Fourier transform imaging states the relationship directly: reducing the pixel raster size by twofold, for example from 50 to 25 microns, results in fourfold more mass spectra collected. That is the whole of it, and its consequences are severe enough to deserve arithmetic rather than description.
Pixel Size | Pixels in 10 x 10 mm | At 1 spectrum/s | At 10/s | At 50/s |
50 microns | 40,000 | 11.1 hours | 1.1 hours | 0.2 hours |
25 microns | 160,000 | 44.4 hours | 4.4 hours | 0.9 hours |
10 microns | 1,000,000 | 11.6 days | 27.8 hours | 5.6 hours |
5 microns | 4,000,000 | 46.3 days | 4.6 days | 22.2 hours |
Table 2. Acquisition time for a 10 by 10 mm area, calculated from pixel count and acquisition rate alone. Excludes stage movement, changeover, and any per-pixel overhead, so treat these as lower bounds rather than estimates.
Read the Grid in Two Dimensions Moving from 50 micron to 5 micron pixels multiplies pixel count by exactly 100. That single factor explains why most published biological imaging work sits between 20 and 50 microns regardless of what the instruments can achieve, and it is the reason the SIMS resolution figure discussed in SIMS and High-Resolution Elemental Imaging is achievable but rarely reported. The columns matter as much as the rows. Analyser choice sets your acquisition rate: Fourier transform analysers operating at extreme resolving power sit near the left of this table, while time-of-flight platforms sit near the right. So the two instrument decisions are not independent. An FT analyser at maximum resolving power combined with a 5 micron raster is not two defensible preferences — it is a combination that produces a month-long acquisition on a single section. Check the intersection before committing to either choice. |
Two refinements make the grid more useful. Shots per pixel multiply the figures further, since building adequate signal at a position often requires summing many laser shots, so a table computed at one spectrum per pixel understates a real MALDI experiment. And imaged area is the term most easily reduced: at 10 microns, a 2 by 2 mm region of interest is 40,000 pixels, the same count as a whole 10 by 10 mm section at 50 microns. Choosing a region rather than a section is frequently the difference between a feasible high-resolution experiment and an abandoned one.
What Actually Limits You?
Diagnose before optimising, because the three trade-offs have different fixes and effort spent on the wrong one is wasted. Four questions identify the binding constraint.
- Does signal disappear, or does the image just look coarse? If your target analyte is detected but the image lacks structural detail, you are resolution-limited and the fix is spatial. If the target vanishes at finer pixels, you are sensitivity-limited.
- Does the number of detected features fall as you refine? If the target survives but the feature count drops, that is coverage, and the response is chemical rather than spatial.
- Is the acquisition simply too long to run? A time constraint, addressable by area, analyser, or accepting coarser pixels, and it is the constraint most amenable to arithmetic in advance.
- Have you verified crystal or probe geometry? If matrix crystals exceed your intended pixel size, or a probe footprint is wider than the raster step, the nominal resolution is not being achieved regardless of instrument settings.
The fourth question catches a surprisingly common situation: an experiment set to a fine raster step that is not actually resolving at that scale, because the physical sampling unit is larger than the step. Oversampling of that kind can improve apparent image smoothness while adding acquisition time and no information. Verifying crystal size or probe footprint takes little effort and settles it.
Strategies That Buy Back Resolution or Sensitivity
Five approaches recover something without simply accepting the loss. None is free, and knowing what each costs is the point.
Strategy | What It Recovers | What It Costs |
More shots or longer dwell per pixel | Sensitivity at fixed resolution | Acquisition time, multiplied across every pixel |
Finer matrix crystals or tighter probe geometry | Genuine resolution, removing a physical ceiling | For sublimation, some analyte coverage; for probes, alignment effort |
Higher mass resolving power | Confident assignment in a crowded spectrum | Acquisition rate, which pushes you leftward in Table 2 |
Ion mobility separation | Ability to separate isobaric species | Little acquisition time, which makes it unusually attractive |
Reduce imaged area to a region of interest | Feasibility at fine resolution | Spatial context, and the risk of sampling bias |
Computational image fusion with microscopy | Apparent spatial detail beyond acquired resolution | It is a predictive result, not a measurement, and must be reported as such |
Table 3. Strategies and their costs. The fourth row is the closest thing to a free improvement available; the last must be described accurately or it misleads.
Ion mobility deserves the emphasis. Because it separates ions by collision cross-section alongside mass, it addresses the isobaric overlap problem that dominates lipid work, and it does so without slowing the raster appreciably. A study of ultrahigh-resolution imaging reports a MALDI quadrupole time-of-flight instrument with trapped ion mobility achieving an increase in peak capacity of more than 250 percent during ion mobility experiments. Gaining separation without paying in time is rare in this field.
The final row needs care in reporting. Fusing lower-resolution ion images with high-resolution optical microscopy can produce images at apparent resolutions finer than anything acquired, and that is a legitimate and useful technique. It is also a prediction informed by two measurements rather than a measurement itself, and the distinction matters when a figure is quoted. The same caution applies to the resolution figures discussed in DESI and Ambient Ionization Imaging, where fusion results and acquired results are routinely conflated in secondary coverage.
What Pixel Size Should You Use?
Work from the biological structure you need to resolve, not from the instrument specification. Four steps.
- State the smallest feature that matters. If the question concerns tissue regions, tens of microns will do. If it concerns individual cells, you need pixels smaller than a cell. If it concerns subcellular compartments, you need a different technique.
- Sample at two to three pixels across that feature. A feature the same size as one pixel is not resolved in any useful sense. This step usually reveals that the required resolution is coarser than instinct suggested.
- Check the arithmetic against Table 2. Compute pixel count for your actual area and divide by a realistic acquisition rate. If the answer is days, revise the area or the resolution now rather than after a failed run.
- Verify coverage at that setting on your own tissue. Run a small test region and count detected features, rather than assuming that coverage established at coarser pixels carries over.
If the Question Is | A Reasonable Starting Point |
Which tissue regions differ | 50 to 100 microns. Fast, sensitive, and sufficient |
Where a structure boundary lies | 20 to 50 microns, with a defined region of interest rather than a whole section |
Cell-level localisation | 10 microns or finer, accepting reduced coverage and a long acquisition |
Subcellular localisation | A different technique. This is where SIMS applies |
Broad untargeted survey | Coarser than instinct suggests. Coverage is the objective, not detail |
Table 4. Starting points by question type. These are opening positions for optimisation on your own tissue, not recommendations.
The last row is worth dwelling on because it inverts a common instinct. In an untargeted survey, the objective is to see as many species as possible, and finer pixels actively work against that. Choosing coarse pixels deliberately for a discovery experiment, then imaging a selected region finely once you know what matters, is a more productive sequence than attempting both at once.
For where these constraints sit within the broader spatial landscape and how mass spectrometry compares with antibody and sequencing-based techniques, see Spatial Analysis in Analytical Science: Mass Spectrometry Imaging and Spatial Omics.
This article was produced under Separation Science's AI Editorial Guidelines.
