I recently watched the excellent training video for the LSM880 by Jason Kirk for reference of our own facility. In his video he mentions using the Digital Offset slider for the PMT to eliminate “blue” (underexposed?) pixels. I always thought that I should not touch this setting, as I was afraid I would be cutting off “real” background pixels, that would better be subtracted during post-processing.
A previous thread discussed this a bit, but largely focused on saturation. I’d like to open up the discussion on “what should be done” for underexposed pixels. Is it bad to leave pixels underexposed for quantification later?
For the purposes of quantitation, all the same concepts from saturated pixels apply for “undersaturated” pixels. Though it’s not uncommon advice, there’s almost never a good reason to have pixels with an intensity value of zero in your image* (outside of the desire to make the image look “pretty” … which you should do in post as you said). That doesn’t mean you shouldn’t touch the offset setting (you may need to adjust it to maximize your dynamic range… that’s why they give you the control in the first place)…
just don’t abuse it for qualitative reasons.
*a possible exception to this are detectors that are basically always in photon-counting mode, in which case you may likely see lots of zero-valued pixels
This is precisely why you should adjust this setting. If you ignore it then there is a strong possibility that you could have large sections of zero grey level pixels in your image (colored blue with the range indicator). The problem with this is that you don’t know if they are actually zero or are somewhere below zero because images don’t record negative values.
It is important to understand what this Digital Offset value is doing on commercial microscopes. The vendor (Zeiss in this case) is calibrating the digital offset slider so that when this value is set to 0 – you should be able to scan an image with no light (no lasers on – just ambient light conditions) and have the baseline grey level for each pixel be just above 0. In the case where the vendor calibration is perfect – you should not need to adjust it further.
This is where you need to take vendor instruction with a grain of salt – because they will all tell you their systems are always calibrated perfectly.
Users must understand that every system is different – and do not have the same calibrations as another. Because of this – you may encounter situations where when this slider is set to 0, you have massive areas of zero level pixels. This is especially true if the calibration is not done well – or has not been done for quite some time.
Now does this mean we should rush to ‘calibrate’ the offset? Not necessarily – IF you know how to balance the offset as we discuss in our training video – you do not care what the calibration is. And as it can fluctuate with temperature, humidity, and time – it will always need some tweaking for best results.
Remember that you are collecting raw data at this stage - so as @talley stated - you must treat the low end of the dynamic range just as you treat the high end by not having any over-exposed or under-exposed pixels in your final image.
Thank you for the great insight. Indeed I did not realize that maybe I would potentially already be cutting off background signal even if the digital offset is set to 0. I will make sure future users are more aware of these settings (including myself!)
@jasonkirk I agree it is a bit misleading when a vendor describes a system as “calibrated.”
So it is possible to calibrate lets say a multimeter, oscilloscope, spectrometer or a laser power meter. This is due to the fact that there are internationally recognized reference standards and a chain of traceability back to those. For some scientific instruments it is easy to obtain, a standardized NIST reference sample of fluorescein for calibration of a detector. However, it is not an imaging instrument.
In the machine vision world, it is possible to calibrate a camera and the resulting image on a display according to some standard test targets.
However, there are not any well-defined standards for “calibration” as it pertains to confocal or fluorescence microscopy in general. ISO is working on generalized interim standards. In the general sense of scientific intstruments…it is not actually possible to “calibrate” a confocal microscope.
Many companies will use some test slides from Invitrogen/Thermofisher (Fluocells) or GattaQuant (GattaCells) to show their system to customers or for QC.
But the problem is that the “reference” samples are not stable over time, not standardized, and subject to minor sample-to-sample variations (non reproducible) amongst a multitude of variables.
The Handbook of Biological Confocal Microscopy Chapter 36, Practical Confocal Microscopy, pp. 660-662 (the 3rd edition I have on hand) the sections under Signal Detection including Signal-to-Noise Ratio and Coeffcient of Variation, Instrument Dark Noise, Photon (Shot) Noise, and PMT Linearity may make for some interesting background reading with regards to a single detection channel. There are various suggestions for non-biological samples to try as reference samples (mostly similar to things people have readily available). Nothing, other than asking an advisor for time on a public system.
