Real-time GPU-based 3D Deconvolution

Abstract

Abstract:Confocal microscopy is an oft-used technique in biology. Deconvolution of 3D images reduces blurring from out-of-focus light and enables quantitative analyses, but existing software for deconvolution is slow and expensive. We present a parallelized software method that runs within ImageJ and deconvolves 3D images ~100 times faster than conventional software (few seconds per image) by running on a low-cost graphics processor board (GPU). We demonstrate the utility of this software by analyzing microclusters of T cell receptors in the immunological synapse of a CD4 + T cell and dendritic cell. This software provides a lowcost and rapid way to improve the accuracy of 3D microscopic images obtained by any method.

CINCH (confocal incoherent correlation holography) super resolution fluorescence microscopy based upon FINCH (Fresnel incoherent correlation holography)

“…Building on the Microvolution (Danville, CA) technique of GPU-based deconvolution 22 implemented as a plug-in to ImageJ 23 , we developed a blind Richardson Lucy-type deconvolution routine to process the complex valued CINCH images. For the complex deconvolution, done here on each color channel as a two-dimensional deconvolution of a single focused reconstructed image, an empirical point spread function (PSF) was created as a starting point by acquiring holograms of subresolution beads, propagating the holograms to reconstruct the focused image of the subresolution beads, cropping out single beads, registering the beads to a constant reference, and finally averaging the beads.…”

“…For the confocal images, we performed 7 iterations of a blind Richardson Lucy deconvolution algorithm. The starting PSFs were created by a theoretical Fraunhofer diffraction model 22 with a numerical aperture of 1.49, refractive index of 1.515, backprojected pinhole radius of 778 nm, and emission wavelength of 525 nm or 590 nm for the golgi or microtubule images, respectively.…”

Using Fluorescence Resonance Energy Transfer-Based Biosensors to Probe Rho GTPase Activation During Phagocytosis

“…An identical set of raw input images presented in Fig. 1 was processed for deconvolution using the Microvolution GPU-deconvolution package [28]. Image processing and ratiometric calculations were performed as in Fig.…”

“…These newer GPU cards, which are quite affordable at only a few hundred dollars, contain thousands of parallel computing processing units to accelerate the video graphics capability of a computer. A new deconvolution software package from Microvolution LLC [28] is now taking advantage of GPUs to accelerate the deconvolution calculations by several orders of magnitudes. This package was used to deconvolve the same Z-series used in Fig.…”

Enhanced 3D localization of individual RNA transcripts via astigmatic imaging

“…Figure 4 is a three-dimensional voxel representation of detected RunX2 mRNA in confluent MSCs 14 days post stimulation for both wide-field (green spheres) and astigmatic (red spheres) in the same image area. The nuclei are shown are blue and deconvolved using a GPU based algorithm 35 and we have omitted the cellular membrane for ease of visualization because the MSCs have already formed a threedimensional matrix by 14 days. Using astigmatic imaging, we are able to detect more RunX2 mRNA with higher axial resolution.…”