Optimizing Image Analysis for Subwavelength Fluorescence Microscopy with Palm and Storm

Abstract

Emerging super-resolution fluorescence microscopy techniques (e.g. PALM and STORM) are of growing significance in biophysical research, enabling high resolution imaging of live cells. Key structures imaged by these techniques include the cytoskeleton, membranes, and mitochondria. Recent theoretical work confirms that the experimentally achievable image acquisition rate and resolution in these techniques is limited by the performance of the rejection algorithm (used to distinguish single-fluorophore images from multi-fluorophore images) as much as by the physical performance of the imaging system. Better rejection algorithms may therefore yield faster and more accurate experiments as well as faster post-processing.We benchmarked the performance of several shape-based rejection algorithms that require no a priori knowledge about the fluorescence efficiency or orientation of the probes, as these parameters are subject to considerable variation. We initially characterized an approach to rejection based on a process of (1) nonlinear curve fitting of the intensity map to an asymmetric Gaussian and (2) subsequent rejection or acceptance of images based on the ellipticity of the fitted function. Ellipticity is used to indicate the presence of multiple activated fluorophores that are separated by less than the wavelength of light and forming overlapping blurs with different centers. We found that the minimum separation for reliable rejection was approximately lambda/3. We then characterized an iterative noise-compensated linear curve-fitting algorithm and found its rejection performance to be nearly identical to the nonlinear approach, but significantly faster. Additionally, we have preliminary performance data for a novel rejection algorithm that employs center of mass estimation on different portions of the bright spot to infer ellipticity. These results are promising steps towards STORM/PALM image processing tools fast enough to enable real-time (rather than post facto) visualization of live cells during experimental manipulations.