Conventional flow cytometry is very good at one thing: measuring how much of a signal each cell carries, thousands of cells per second, across many parameters at once. That strength is also its boundary. A photomultiplier reports an integrated intensity, not a location. So the moment your question shifts from “how much” to “where,” the dot plot goes quiet.
This post is about recognizing that moment. Not every assay needs pixels. But a specific set of questions in immunology and cell biology cannot be answered by intensity alone, and forcing them onto a standard analyzer produces data that looks clean and means less than it appears to.
In This Article
What a dot plot cannot show you
Picture a double-positive population in the corner of a two-color plot. Two interpretations fit the same dots: single cells genuinely co-expressing both markers, or doublets where a positive cell is stuck to another cell. Gating on area and height helps, and it is worth doing, but it never fully closes the gap. You are inferring morphology from a pulse shape.
The same limitation shows up whenever biology depends on position. A transcription factor sitting in the cytoplasm and the same factor concentrated in the nucleus can give you similar total fluorescence. A receptor on the membrane and a receptor internalized into vesicles can integrate to the same number. Conventional flow sees the sum. It does not see the arrangement.
The categories of question that call for per-cell images
It helps to sort these by the kind of spatial information you actually need. A few recurring categories:
Translocation. Has a signal moved from one compartment to another, for example NF-κB moving from cytoplasm to nucleus after stimulation.
Internalization. Has a surface receptor been pulled inside the cell, and to what degree across the population.
Co-localization. Do two probes occupy the same space within the cell, or do they merely appear in the same cell.
Cell-cell interaction. Are two cells forming a contact, such as a T cell and an antigen-presenting cell building an immune synapse, with markers polarizing toward the contact site.
Morphology-based classification. Cell-cycle stage read from the nuclear image and DNA distribution, or telling intact cells apart from debris and aggregates.
Image-confirmed rare events. A phenotype rare enough that you want a picture of each hit before you believe it.
Assays where imaging flow earns its place
Take NF-κB translocation. On a plate reader you get a well average; on a microscope you get beautiful images of a few hundred cells. Imaging flow sits between them, scoring nuclear localization on tens of thousands of single cells with a similarity feature that compares the transcription factor image to the nuclear image. You get a distribution, not an anecdote, and you can pull up the cells behind each tail.
Receptor internalization works the same way, using a feature that contrasts membrane signal with interior signal, so you can quantify the shift across a treatment series rather than eyeball a handful of fields. For the doublet problem, the readout is simply the brightfield image: a real double-positive is one cell, and a conjugate is two. Where a lab runs work like this routinely, an imaging flow cytometry platform such as the ImageStreamX Mark II records brightfield, darkfield, and fluorescence images for every event, which is what makes these features possible. Merkel Technologies, which represents the instrument in Israel, can run a proof-of-concept on your own samples before purchase, and that trial usually tells you more than any feature list.
Panels, controls, and the reality of your sample
Adding imagery does not remove the ordinary discipline of flow. Compensation still matters. Titration still matters. What changes is that some controls become image controls. A translocation assay needs a genuinely unstimulated population to define the cytoplasmic baseline and a strong positive to define full nuclear localization, because the score is only as good as its two anchors. Co-localization needs single-stained controls so you can rule out bleed-through masquerading as overlap.
Sample reality also asserts itself. Imaging rewards clean, well-dissociated single-cell suspensions, and it punishes clumps more visibly than a conventional analyzer does, because you can actually see them. Throughput is lower than a standard analyzer and analysis takes longer, so panels and event targets are worth planning with that in mind rather than treating them as an afterthought.
When conventional flow is still the better call
If your question is really about frequencies and expression levels, immunophenotyping a panel, counting a subset, tracking a marker’s median intensity, then imaging flow is the wrong tool. It is slower, the datasets are larger, and the analysis asks more of you. Paying that cost buys nothing when the number alone answers the question.
The honest test is one sentence: does the interpretation of your result depend on where a signal sits, or on whether two things are the same object. If yes, reach for images. If no, a well-run analyzer will serve you better and faster. When you are genuinely unsure, the cheapest way to find out is to run both on a pilot sample and compare what each one actually lets you conclude.
