Cameras for Spatial Transcriptomics and In Situ Imaging Instruments

November 18, 2026

Spatial transcriptomics instruments image the same piece of tissue or the same cells again and again. Each round labels a different set of RNA targets, the instrument images every field, the probes are removed or swapped, and the next round begins. Software then lines up the fluorescent spots from every round and reads the on/off or color pattern at each spot as a barcode for one RNA species.

The approach scales quickly. In sequential FISH, the number of possible barcodes grows as the number of dyes raised to the number of rounds, so four dyes over eight rounds give 65,536 codes (Lubeck et al., Nature Methods, 2014). MERFISH used error-robust codes to measure 140 RNA species in 16 rounds, or 1,001 species in 14 rounds with error detection only (Chen et al., Science, 2015).

For the camera, that means a dim, repeated, highly quantitative measurement. This post covers what to look for, using Finger Lakes Instrumentation (FLI) Revelation cameras as examples. For the general OEM relationship, see our post on choosing an OEM camera partner for life-science instruments.

What multi-round in situ imaging asks of a camera

A single missed or misread spot in one round can turn a correct barcode into an error or a dropout. So the camera has to do three things well: resolve small, dim spots; give the same answer for the same signal in round 1 and round 16; and keep up with a large number of fields, focal planes and rounds without becoming the bottleneck.

Spots, magnification and pixel size

A single-molecule FISH spot is close to the diffraction limit. With far-red dyes around 670 nm and a high-NA oil objective (NA 1.3 to 1.4), the spot's full width at half maximum is roughly 0.25 µm.

Pixel size at the sample is the pixel size divided by the magnification. Revelation cameras use 3.76 µm pixels, so at 40x each pixel covers about 0.094 µm of the sample, or roughly two to three pixels across each spot. That is enough to find spot centers well. At 60x, each pixel covers about 0.063 µm, so the same spot spreads over about twice as many pixels. Finer sampling can help separate crowded spots, but it divides each spot's photons among more pixels, each with its own read noise. Our post on pixel size and sampling covers the math.

Photons per spot and the noise floor

Spot brightness is limited by probe chemistry and by how much light the sample can take across many rounds. Every extra photon spent on one round is bleaching budget taken from the later ones, so exposures are kept short and excitation modest.

That puts weight on the sensor. Back-illuminated sensors collect more of the light that arrives, and the Sony back-illuminated CMOS sensors in Revelation cameras have very low read noise, so faint spots stand above the noise floor without long exposures. Cooling keeps dark signal low enough that it does not set the noise floor.

One detail matters for the noise budget: on these Sony sensors, binning is done digitally rather than on the sensor, so binning does not lower read noise the way on-chip CCD binning does. Choose magnification to get the sampling you want rather than relying on binning later.

Same field, many rounds

Decoding only works if a spot in round 1 lands on the same spot in round 16. Fixed transcripts stay in place between rounds, so the spots can be aligned and read in sequence (Lubeck et al., 2014). In practice, instruments image fiducial beads or a nuclear stain each round and register every image to them.

The camera affects how much correction is needed. Mount it rigidly to the optical path, so that fluid exchange, stage moves and temperature changes do not shift it. Let the stage settle before each exposure, especially with a rolling shutter, where rows are exposed at slightly different times. If you pulse the excitation, the rolling-with-global-reset mode lets every row see the same light pulse. Our post on automated fluorescence imaging instruments covers that timing in detail.

Keeping every round comparable

Decoding compares intensities across rounds, so anything that drifts looks like biology. Hold the sensor temperature, gain mode, readout mode and exposure fixed for the whole run. Take flat fields for each channel, and take dark frames by switching off the excitation or moving to a blocked filter position. Revelation and Aurora cameras are not offered with a mechanical shutter, and these methods suit an automated calibration routine.

Across a fleet of instruments, unit-to-unit consistency matters too. Data-sheet values are typical, and a production test summary exists for each camera by serial number. Our post on acceptance testing shows what to check on arrival.

Tiles, z-planes and the data budget

At 40x, the 23.5 × 15.7 mm sensor in the RV571 covers about 0.59 × 0.39 mm of the sample. The 36 × 24 mm sensor in the RV455 covers about 0.90 × 0.60 mm, so a 5 × 5 mm region takes roughly 46 tiles instead of about 108, before overlap. Fewer tiles mean fewer stage moves, focus steps and filter changes in every round. A larger sensor only helps if the optics deliver a flat, sharp field across it. The 16 mm diagonal of the RV533 fits most microscope fields easily, while the larger sensors need wide-field optics.

Data volume is set by area and pixel size, not by sensor format. Imaging that 25 mm² region at five focal planes, two channels and 16 rounds produces about 0.9 TB of 16-bit images with either camera. On-camera memory is a working frame buffer, so plan for continuous transfer to the host and for storage and analysis to match.

Filter changes add up across rounds. The FLI High Speed Filter Wheel moves to the adjacent filter in 23 ms on the 6-position model and 30 ms on the 10-position model. See our post on fast filter switching.

Fitting the camera into the instrument

Revelation puts the same Sony sensors as Aurora into a thin, fin-less body that is shorter front to back. It can be liquid-cooled or mounted to your housing as a heatsink, which helps when the instrument housing is closed and its temperature matters. See our post on air vs liquid cooling.

External trigger in and out are standard, so the camera can follow the instrument's timing controller. Every Revelation camera includes our software development kit and support for integration into your OEM design. Drivers and manuals are on our support page. Start with a standard camera; if your housing needs a small mechanical change, ask our sales team.

Questions to settle before you choose

  1. What magnification and NA set your spot size and pixel size at the sample?
  2. How bright is your dimmest spot, and how much exposure can the sample take across all rounds?
  3. How many rounds, channels and focal planes per run?
  4. How will you register rounds, and how stable is the camera mount?
  5. How much area per run, and how much data can your host store and process?
  6. How much space, airflow or coolant is available inside the enclosure?

Talk to our sales team

FLI cameras are Made in USA, and our sensor chambers are assembled in our 1,000+ sq ft ISO Class 7 clean room. See how on our in-house manufacturing page.

Send our sales team your optical layout, round and channel plan, and throughput target, and we'll help you choose a camera and filter wheel for your instrument. Request a quote or email sales@flicamera.com.