Microtubule tracking from stochastic optical reconstruction microscopy images

TL;DR

This paper presents a computational method for tracking microtubules in super-resolution microscopy images, enhancing the ability to quantify sub-cellular structures despite noise limitations.

Contribution

The work introduces a novel combination of image processing and Markovian algorithms to automatically extract microtubule information from stochastic optical reconstruction microscopy images.

Findings

Effective noise reduction in high-resolution images

Accurate microtubule localization and tracking

Enhanced quantification of sub-cellular structures

Abstract

Our work aims at using quantitative imaging tools to complement the limitation of noise encountered by high resolution fluorescence microscopy methods. Several cycles of fluorophore activation, imaging and deactivation produce a sequence of images in which the signals of individual fluorophores do not overlap, due to the low light intensity during their activation. The centroid position of each fluorophore is then determined by Gaussian fitting of each signal, where the final resolution depends on the precision with which each fluorphore is localized. Superimposing the images will result in having the same fluorophore mapped onto a `cloud' of locations. The most significant information of the superimposed images is contained in the macro-structures identifying microtubules, mitochondria or other organelles. Cascades of binary image processing algorithms are applied in order to isolate the larger organelles. A Markovian algorithm selecting the nearest neighbour is finally applied to the de-noised images, to automatically extract relevant information on microtubules. Our work supplements advancements in experimental technologies with computational methods, helping quantifying sub-cellular properties with high accuracy.