A machine learning approach for online automated optimization of super-resolution optical microscopy

Audrey Durand,Marc-André Gardner,Flavie Lavoie-Cardinal,Christian Gagné,Louis-Émile Robitaille,Theresa Wiesner,Paul De Koninck,Anthony Bilodeau

doi:10.1038/s41467-018-07668-y

Audrey Durand, Marc-André Gardner + Show 6 more

Open Access

https://doi.org/10.1038/s41467-018-07668-y

Copy DOI

Journal: Nature Communications	Publication Date: Dec 1, 2018
Citations: 45	License type: open-access

Affiliation: Université Laval

Abstract

Traditional approaches for finding well-performing parameterizations of complex imaging systems, such as super-resolution microscopes rely on an extensive exploration phase over the illumination and acquisition settings, prior to the imaging task. This strategy suffers from several issues: it requires a large amount of parameter configurations to be evaluated, it leads to discrepancies between well-performing parameters in the exploration phase and imaging task, and it results in a waste of time and resources given that optimization and final imaging tasks are conducted separately. Here we show that a fully automated, machine learning-based system can conduct imaging parameter optimization toward a trade-off between several objectives, simultaneously to the imaging task. Its potential is highlighted on various imaging tasks, such as live-cell and multicolor imaging and multimodal optimization. This online optimization routine can be integrated to various imaging systems to increase accessibility, optimize performance and improve overall imaging quality.

Highlights

Traditional approaches for finding well-performing parameterizations of complex imaging systems, such as super-resolution microscopes rely on an extensive exploration phase over the illumination and acquisition settings, prior to the imaging task
To quantify the success of the imaging process, we evaluated different measurements that were compatible with online analysis on a standard computer: the autocorrelation amplitude of the actin periodical pattern[17], the SNR, the photobleaching, and the Fourier Ring Correlation (FRC)[18,19] (Fig. 1b and Supplementary Figures 2 & 3)
We combined it with neural network approaches for image quality recognition and user preference evaluation to develop a fully automated optimization platform for super-resolution microscopy

Summary

Introduction

Traditional approaches for finding well-performing parameterizations of complex imaging systems, such as super-resolution microscopes rely on an extensive exploration phase over the illumination and acquisition settings, prior to the imaging task. Point scanning methods such as STimulated Emission Depletion (STED)[3] and REversible Saturable Optical Linear Fluorescence Transition (RESOLFT)[4] are well suited for the study of dynamical processes in living cells or multicolor imaging[5,6,7] As they are further developed, along with the continuous evolution of bio-imaging tools and experimental paradigms, these super-resolution methods come with several layers of complexity in their implementation in the laboratory[8,9,10]. We show that complex techniques such as STED, multimodal microscopy, or live-cell optical nanoscopy strongly benefit from a fully automated parameter optimization This strategy could be implemented on a wide range of microscopes, without restriction to super-resolution, to conduct the optimization simultaneously with the imaging task. This strategy should increase the efficiency of the imaging process and standardize the results

Objectives

Methods

Results

Conclusion