Camera parameters for single molecule localization microscopy

Image Analysis
,
#1

Hello, I’m analyzing STORM data and wonder how experts determine the camera parameters (e.g. electrons per A/D count, actual EM gain, etc) to achieve the best possible single-molecule localization precision. Below are some questions specific to the instruments/software I’m using, but any general advice and reference on this topic would be greatly appreciated!

  1. What’s the best practice for determining or measuring the parameters?

  2. I’m using Nikon Elements software and Andor iXon Ultra 897 EMCCD. Andor’s manual says their EMCCDs feature “RealGain”, which enables the user to select actual gain directly from a linear scale displayed in software. Is this feature automatically enabled when using Nikon Elements? If not, how can enable/disable this feature?

  3. Same question for “Offset clamp” feature of Andor iXon 897 EMCCD. Is this feature enabled by default? How can I control this?

  4. How reliable are those RealGain and Offset clamp? Has anyone compared the measured gain to the gain selected in software when RealGain feature is on?

  5. In your experience, or theoretically, how much does having camera parameters accurately measured or estimated based on spec sheets and software affect the end results?

#2

Hello,

  1. I’m not sure what is considered best practice, but two there are two ways to do this that I am aware of. The first way is to turn off the EM Gain and calibrate the CCD using the fact that the noise from a light source is Poisson. You record movies at different light intensities, then plot the mean versus variance at each intensity to determine the A/D response. Then once you know the A/D response you can measure the EM gain response by varying the EM gain while imaging a constant light source. On a microscope I usually use the bright field lamp, which is good enough for this purpose. The second way is to just assume that the EM gain stage only adds the theoritically expected amount of additional noise (an increase in variance of a factor of 2) and calibrate following step 1 of the first way, but including the factor of 2 when fitting the mean vs variance plot.

  2. Sorry I’m not familiar with Nikon Elements.

  3. Same as (1).

  4. I believe the RealGain is fairly reliable. Your measured gain values will probably be within 10% of the RealGain value. This will also be true of A/D conversion values (pre-amp gain) which Andor provides, or at least used to provide. It has been a while and I don’t remember what Offset clamp does for you.

  5. This depends on the analysis software you are using. Some analysis packages assume that the image data is Poisson to segment the data. They have a threshold that is signal dependent, so if the signal has an unexpected scale they may do a poor job segmenting. For fitting, again it depends, analysis software that is using least-squares fitting, or some variant such as weighted least-squares won’t be sensitive to the camera parameters. In theory analysis software that is using MLE fitting would be sensitive, but I believe that you have to be way off, like an order of magnitude, to see any noticeable effect. Probably better to test this though, instead of taking my word for it…

1 Like
#3

Thanks for your response!