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

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.

Enhancing neuronal chloride extrusion rescues α2/α3 GABAA-mediated analgesia in neuropathic pain

“…Enhanced resolution microscopy using multi-array Airyscan. While superresolution approaches 129 have been very instrumental to decipher biological structures at the nanoscale level in some special circumstances (i.e., isolated cells, brain slices) [130][131][132] current super-resolution techniques still do not effectively provide significant improvements in resolution in sections of the thickness normally used for immunocytochemistry particularly if extensive field of view need to be studied 129 . This is because improvements in spatial resolution always come in pair with sacrificing image speed acquisitionhence ability to scan large fields of view-and fluorophore flexibility (i.e., very long scan times causing photobleaching limitations).…”

Single-shot 3D wide-field fluorescence imaging with a Computational Miniature Mesoscope

“…Second, the imaging optics of the CM 2 is designed by heuristically balancing several hardware and imaging attributes, including the resolution, FOV, image contrast, and device complexity and size. Given this multidimensional design space and several intrinsic trade-offs, it is highly possible that the imaging optics can be further optimized by using advanced computational procedures, such as those based on classical [e.g., the genetic algorithm ( 34 )] or data-driven [e.g., machine learning ( 35 , 36 )] algorithms. Here, we discuss several promising directions to pursue in the future.…”

“…Last, the image contrast is primarily affected by the array size of the multifocus PSF and the scattering conditions. It may be possible to improve the contrast by reducing the array size while maintaining the imaging resolution and FOV by optimizing the physical parameters of the MLA using advanced algorithms, such as the genetic algorithm ( 34 ) and deep learning ( 35 , 36 ). In addition, advancing the illumination technology by incorporating the structured illumination ( 43 ) can be a promising solution to suppress the background fluorescence and improve the image contrast.…”

A Review of Intrinsic Optical Imaging Serial Blockface Histology (ICI-SBH) for Whole Rodent Brain Imaging

“…Similar advances were presented in the field of augmented reality microscopy. With the use of modern computer vision and machine learning methods, these novel systems can perform automated cell tracking with a convolutional neural network trained for object detection [161], cancerous tissue identification with image classification networks [162], or the use of machine learning for online automated optimization of microscopy [163].…”