Correcting Artifacts in Single Molecule Localization Microscopy Analysis Arising from Pixel Quantum Efficiency Differences in sCMOS Cameras

Hazen P Babcock,Fang Huang,Colenso M Speer

doi:10.1038/s41598-019-53698-x

Abstract

Optimal analysis of single molecule localization microscopy (SMLM) data acquired with a scientific Complementary Metal-Oxide-Semiconductor (sCMOS) camera relies on statistical compensation for its pixel-dependent gain, offset and readout noise. In this work we show that it is also necessary to compensate for differences in the relative quantum efficiency (RQE) of each pixel. We found differences in RQE on the order of 4% in our tested sCMOS sensors. These differences were large enough to have a noticeable effect on analysis algorithm results, as seen both in simulations and biological imaging data. We discuss how the RQE differences manifest themselves in the analysis results and present the modifications to the Poisson maximum likelihood estimation (MLE) sCMOS analysis algorithm that are needed to correct for the RQE differences.

Highlights

This work discusses the corrections necessary to the Poisson maximum likelihood estimation (MLE) scientific Complementary Metal-Oxide-Semiconductor (sCMOS) algorithm described in Huang et al.[6] to handle pixel-dependent relative quantum efficiency (RQE) differences
We found that the “RQE Correction” fitting approach is an unbiased estimator of the ground truth emitter position, with bias values that are within the estimated error of zero (Fig. 3B,C)
When the analysis was not corrected for pixel-to-pixel RQE differences, the probability the correlation was random was less than 10-13, but when RQE correction was included the probability that the correlation was random increased to ~0.2. This indicates that the corrections we propose are successfully compensating for pixel-dependent RQE differences

Summary

Introduction

This work discusses the corrections necessary to the Poisson MLE sCMOS algorithm described in Huang et al.[6] to handle pixel-dependent RQE differences. Other labs have presented work that discusses correcting for the pixel-dependent differences in offset, gain, variance and quantum efficiency (QE) for sCMOS cameras for optimal www.nature.com/scientificreports weighted least squares fitting[11,12] In these works the authors use the average camera gain as the gain value for each pixel. If the camera has substantial pixel level differences in QE the result from using the average gain and a flat field correction term to convert camera values (ADU) to photo-electrons (e-) will no longer be a Poisson distribution with mean equal to the variance Because this can degrade fitting performance when the fitter assumes Poisson statistics, we instead use a pixel-dependent gain value that preserves the expected relation between mean and variance and a pixel-dependent RQE term to compensate for any differences in pixel QE. We focus on this problem as it has been shown that weighted least squares fitting is not as accurate as MLE fitting on Poisson distributed data, especially at the lowest signal levels[13,14]

Methods

Results

Conclusion