Highly Efficient and Scalable Framework for High-Speed Super-Resolution Microscopy

Quan Do,Krishna Agarwal,Sebastian Acuna,Jon Ivar Kristiansen,Phuong Hoai Ha

doi:10.1109/access.2021.3094840

Quan Do, Krishna Agarwal + Show 3 more

Open Access

https://doi.org/10.1109/access.2021.3094840

Copy DOI

Journal: IEEE Access	Publication Date: Jan 1, 2021
Citations: 2	License type: CC BY 4.0

Affiliation: UiT The Arctic University of Norway

Abstract

The multiple signal classification algorithm (MUSICAL) is a statistical super-resolution technique for wide-field fluorescence microscopy. Although MUSICAL has several advantages, such as its high resolution, its low computational performance has limited its exploitation. This paper aims to analyze the performance and scalability of MUSICAL for improving its low computational performance. We first optimize MUSICAL for performance analysis by using the latest high-performance computing libraries and parallel programming techniques. Thereafter, we provide insights into MUSICAL's performance bottlenecks. Based on the insights, we develop a new parallel MUSICAL in C++ using Intel Threading Building Blocks and the Intel Math Kernel Library. Our experimental results show that our new parallel MUSICAL achieves a speed-up of up to 30.36x on a commodity machine with 32 cores with an efficiency of 94.88%. The experimental results also show that our new parallel MUSICAL outperforms the previous versions of MUSICAL in Matlab, Java, and Python by 30.43x, 2.63x, and 1.69x, respectively, on commodity machines.

Highlights

In biomedical imaging, there is a limitation in resolving details smaller than the Abbe diffraction limit, which itself is the radio of the fluorescence emission wavelength to twice the numerical aperture of the microscopy
Our experimental results show that our new parallel multiple signal classification algorithm (MUSICAL) achieves a speed-up of up to 30.36x on a commodity machine with 32 cores
To make MUSICAL more user-friendly and easier to use, we propose an auto-threshold algorithm in step 6

Summary

Introduction

There is a limitation in resolving details smaller than the Abbe diffraction limit, which itself is the radio of the fluorescence emission wavelength to twice the numerical aperture of the microscopy. There are several techniques to overcome the resolution limit, such as singlemolecule localization microscopy (SMLM) [1,2,3,4,5,6], including stochastic optical reconstruction microscopy (STORM) [7], photo-activated localization microscopy (PALM) [8,9], multiple signal classification (MUSIC) [10,11,12], structured illumination microscopy (SIM) [13], and fluorescence fluctuations based super-resolution microscopy (FF-SRM) [14]. The multiple signal classification algorithm (MUSICAL) [15] belongs to the family of fluctuations-based super-resolution microscopy and is the primary focus of this work. Their article included a comment on the potential of parallelizability, which has so far not been investigated further

Objectives

Results

Conclusion