Skip to content

Particle normalization

Michael A. Cianfrocco edited this page Jun 7, 2018 · 1 revision

Home > Best practices > Particle normalization

##Particle normalization

While collecting data on an EM, the recording devices will register counters per incident electron that hits the detector. This means that the **raw pixel values need to be adjusted **between images so that statistical comparisons between the single particles is not confounded by the nature of detector statistics.

Furthermore, changes in defocus will alter the distribution of pixel intensities values, which means that changing the defocus changes the standard deviation of the observed pixel intensity values.

Therefore, image normalization for single particle EM will perform the following operations:

  1. Set the average pixel value to zero
  2. Scale the standard deviation of observed pixel intensities to a set value (1, 3.5, or 10)

Beyond changing the average and standard deviation, programs may provide the following options / choices for normalization:

####Edgemean

When calculating the average and standard deviation of pixel values for a given single particle, programs normally use all pixels across a given image. Instead, for certain circumstances, you can normalize your image using the outer rim of pixels that do not containing protein to get a more accurate sense of the background noise and standard deviation levels.

####Ramp

It is common to observe gradients in stain and ice thickness, which manifest as continuous changes in pixel intensity values across a single particle image. To remove these gradient values, there is a correction called 'Ramp correction' that will make sure to even out the image statistics when there is a gradient across the image.

####Remove outliers

Both Relion and XMIPP provide users with the choice to replace 'hot' or 'dead' pixels with random values found with Poisson noise (-3.5 to +3.5). These pixels are replaced ONLY if they fall outside of a user specified range. Typically, any pixel intensity that is greater than 5 standard deviations away from the average is the result of an X-ray or a dead pixel in the detector.

These outliers can be removed by setting '5' as the limit during Relion or XMIPP normalization.

Clone this wiki locally