Mass Spectrometry Imaging: Five Key Physical Processes from Tissue Sections to Ion Images

The final output of mass spectrometry imaging (MSI) is a colour-coded map of ion distribution. Many users begin biological interpretation immediately upon receiving the data; however, each of the five physical processes involved may affect the reliability of the final image—understanding these processes is not merely a matter of intellectual curiosity, but an essential part of data quality control.

Step 1: Sample Preparation—Section Thickness, Matrix and Fixation

The preparation of tissue sections is the first step in MSI, and also the one most frequently underestimated. Sections of different thicknesses can vary by a factor of several in desorption efficiency: if too thin (<5>20 μm), molecules in the inner layers are difficult to reach via desorption, and the image reflects the distribution at the surface of the section rather than throughout the entire section.

Temperature control during cryosectioning is another variable that is easily overlooked. Tissues sectioned at –20°C exhibit a relatively uniform surface morphology; if the sectioning temperature is too high (above –15°C), the tissue softens, resulting in a rough cut surface. At micrometre-level resolution, surface irregularities can cause significant fluctuations in desorption efficiency—differences in signal intensity between two adjacent pixels for the same molecule may stem from concentration variations, or simply from differences in the smoothness of the section’s surface.

The choice of fixation method has a particularly significant impact on lipid imaging. Formaldehyde fixation leads to cross-linking reactions, altering lipid distribution and substantially reducing desorption efficiency. Fresh-frozen sections are the ‘gold standard’ for lipid imaging, but not all laboratories have the facilities for fresh-freezing.

Step 2: Desorption – Transferring molecules from the solid or liquid phase to the gas phase

Desorption is the key determinant of spatial resolution. The sputtering zone diameter for DESI spray desorption is approximately 100–300 μm (depending on the spray capillary diameter, flow rate and nozzle distance), whilst the spot diameter for LDI laser desorption is approximately 2–80 μm (depending on the focusing optical path and laser wavelength).

The chemical selectivity of desorption efficiency is also a variable that cannot be overlooked. DESI is more efficient at desorbing molecules with high surface activity and polarity; LDI is more efficient at desorbing molecules with UV absorption (such as lipids and pigments containing conjugated systems). This implies that desorption alone introduces a chemical bias—a ‘bright spot’ for a particular molecule in the image may reflect both a genuinely high concentration and the fact that the molecule is more readily desorbed in that tissue microenvironment.

Step 3: Ionisation—from molecules to ions

This step has been analysed in detail in previous articles. From the perspective of data quality control, two additional points are worth noting:

Spatial variability in ionisation efficiency. Real tissue sections are not homogeneous matrices; salt concentration, water content and pH vary across different micro-regions, and these differences can significantly affect ESI ionisation efficiency. For the same molecule in different micro-regions of the tissue, ionisation efficiency may vary even if the actual concentration is identical—this directly translates into false-positive ‘concentration differences’ in the image. Internal standard calibration (introducing an isotopic internal standard into the matrix) is a means of mitigating this issue; however, it is difficult to ensure the uniformity of the internal standard at high spatial resolutions (2–5 μm).

Ion suppression across pixels. Between adjacent pixels, if one region contains a large number of molecules with high ionisation efficiency (such as high-abundance PC in neural tissue), the signal from the adjacent low-concentration region may be ‘suppressed’—this is caused by spatial crosstalk from the ion cloud near the mass spectrometer inlet. In DESI imaging, this problem is more pronounced than in LDI because the spatial distribution of the DESI spray plume is broader.

Step 4: Mass Analysis — Ion Flight Time

Time-of-flight mass spectrometry (TOF-MS) is the most commonly used mass analyser in mass spectrometry imaging due to its high full-spectrum acquisition rate (reaching tens to hundreds of spectra per second), which is naturally suited to the point-by-point scanning mode of imaging.

The resolution and mass accuracy of TOF directly determine the chemical specificity of the ion image. At a resolution of R = 5000, two molecules with closely matching mass numbers (such as m/z 760.585 and 760.525, differing by 0.06 Da) may not be completely separated; they merge into a single channel in the image—resulting in an image of ‘chemical mixture’ rather than a distribution map of a single molecule. When the TOF resolution is increased to R > 15,000, these two molecules can be clearly separated, each forming an independent ion image.

Neo Source INSTRUMENT’s ProC series of online time-of-flight mass spectrometers utilise a dual-field reflection TOF design. Through Wiley-McLaren dual-field acceleration and mirror energy compensation, the mass spectral resolution can reach several thousand to over ten thousand, meeting the resolution requirements for most mass spectrometry imaging applications.

Step 5: Data Processing and Visualisation – From Spectra to Images

The final stage of image processing involves extensive data processing: baseline subtraction, peak extraction, peak alignment (correction for m/z drift between different pixels), normalisation and false-colour mapping.

In peak extraction algorithms, the selection of thresholds and signal-to-noise ratio (S/N) cut-off values directly affects the imaging results—low thresholds may omit low-abundance but biologically significant molecules, whilst high thresholds may introduce noise artefacts. The default parameters in most imaging software are not ‘optimal’ but rather ‘moderate’—manually optimising threshold parameters for specific analytical objectives (such as targeting a known metabolite) usually yields significantly better image quality.

Normalisation is another key decision point: TIC (Total Ion Current) normalisation is the simplest but is prone to being dominated by high-abundance ions; internal standard normalisation is more reliable but increases the complexity of sample preparation; median normalisation is statistically robust but may mask true biological gradients. Currently, no single normalisation method is suitable for all scenarios—this is one of the few aspects of MSI data processing that remains highly dependent on the operator’s judgement.

A holistic view of the five steps

These five steps are not independent, sequential processes—errors in each step are amplified in subsequent steps. Section flatness affects desorption consistency; desorption deviations compound spatial variations in ionisation efficiency; and ionisation variability is in turn amplified or suppressed by TOF acquisition parameters… The final ion image is the result of the convolution of these five physical processes.

Understanding the physical limitations of each step is not about finding ‘perfect’ parameters, but rather about ensuring that, when interpreting images, we do not attribute all brightness differences solely to variations in molecular concentration. The gap between ‘what the instrument tells us’ and ‘the true state of the sample’ is the core issue that must be taken seriously in mass spectrometry imaging.

Related products: Neo Source INSTRUMENT MSI DPI (20–200 μm) and MSI LDPI (2–3 μm) mass spectrometry imaging sources, compatible with Agilent, AB SCIEX and Thermo mass spectrometers; ProC series online time-of-flight mass spectrometers, featuring a dual-field reflection TOF design.