Imaging Cytometry
At the recent American Association of Immunologists meeting, I had the time to meet with the Amnis team and learn about their new imaging cytometers the ImageStream and FlowSight instruments. These instruments represent a new class of imaging cytometer and we think instruments like this will provide significant new capability.
Imaging Cytometry: The term “imaging cytometry” has been in use for some time now to describe a wide variety of instruments that study cells by imaging them and processing the resulting image. The Nexcelom Cellometer Vision is a recent example of such an instrument.
After loading a slide with cells, the instrument images them, and then performs analysis of fluorescence intensities (in multiple channels), giving a cell-by-cell plot of these intensities that will be familiar to cytometrists. With easy to load slides, multiple filters, and the ability to identify individual cells (versus background) these systems are quite versatile, and several examples now exist.
Imaging Flow Cytometry: The Amnis instruments are a bit different. They don’t image cells that are spread out on a slide or at the bottom pf a micro titer plate. Instead, these are flow instruments that guide a stream of cells through laser light sources. But, instead of reading fluorescence using a photomultiplier, the signals from these cells are imaged as the cells traverse the light source. So in addition to getting a homogeneous fluorescence intensity (which a conventional cytometer would also capture), the images show the inhomogeneous distribution of the label within the cells. These features can be dissected much like any conventional signal coming from a cytometer. Notice the axes on the figures below.
An Important Step in Model Development: We believe that a key feature of these new cytomters (true of the non-flow imagers was well) is the ability to develop models based on observation and correlation. The Amnis IDEAS software computes a wide variety of morphological features for each cell as it passes through the detector. The features include “circularity” “spottiness” “Internalization”, etc. etc. and reflect characteristics of the spatial distribution of the probe within the cell. A “truth set” can be created and the software can evaluate what features best describe the difference between truth and non-truth. This is an analysis of interclass variance, and the truth sets are typical “treated” and “untreated”, “stimulated” and “unstimulated”. The result is a description of the features that best describe the difference between treated and untreated. This is nicely described in the technical poster on the Amnis site titled Automated Classification of Cellular Images Obtained on the Image Stream Imaging Flow Cytometer. The resulting models are quite powerful, not least because they are unbiased in their construction.
The ability to capture the large quantities of data possible in a flow instrument is critical to gaining the level of statistical significance required to take this model building approach, and we think this will usher in a new era of phenotypic assays.
Why is it on our Radar? Like all modeling efforts, the model is only as good as the data that’s used in its construction. If the instrument is measuring biologically relevant features, then the models will be more significant, and more easily interpreted. That’s important when trying to understand what makes the “treated” and “untreated” states different. We think that targeted biosensors are a natural for this approach, since they allow more and more detailed features to be detected within cells, and these features are connected directly to a biological target, rather than being more abstract, like “circularity.” We believe that the tools we’re developing will be very illuminating when combined with instruments like this, and will enable biologist to get a deeper understanding of what’s happening in a given experiment. And that’s our mission at Sharp Edge.
–The figures in this post are copyright of the individual instrument manufacturers.–
