Data-adaptive image-denoising for detecting and quantifying nanoparticle entry in mucosal tissues through intravital 2-photon microscopy.

Torsten Bölke,Regina Orzekowsky-Schroeder,Andreas Gebert,Horst Weller,Anna Schüth,Antje Klinger,Gereon Hüttmann,Tobias Vossmeyer,Jelena Dimitrijevic,Lisa Krapf

doi:10.3762/bjnano.5.210

Abstract

Intravital 2-photon microscopy of mucosal membranes across which nanoparticles enter the organism typically generates noisy images. Because the noise results from the random statistics of only very few photons detected per pixel, it cannot be avoided by technical means. Fluorescent nanoparticles contained in the tissue may be represented by a few bright pixels which closely resemble the noise structure. We here present a data-adaptive method for digital denoising of datasets obtained by 2-photon microscopy. The algorithm exploits both local and non-local redundancy of the underlying ground-truth signal to reduce noise. Our approach automatically adapts the strength of noise suppression in a data-adaptive way by using a Bayesian network. The results show that the specific adaption to both signal and noise characteristics improves the preservation of fine structures such as nanoparticles while less artefacts were produced as compared to reference algorithms. Our method is applicable to other imaging modalities as well, provided the specific noise characteristics are known and taken into account.

Highlights

Imaging methods applied to detect fluorescent nanoparticles in mucosal tissues should provide high optical resolution and allow large volumes to be scanned
Our method is applicable to other imaging modalities as well, provided the specific noise characteristics are known and taken into account
All of them stated that the modified version we propose led to reduced artefacts and a better preserved representation of nanoparticles and other fine structures compared to the original BM3D algorithm (Figure 7 and Figure 8)

Summary

Introduction

Imaging methods applied to detect fluorescent nanoparticles in mucosal tissues should provide high optical resolution and allow large volumes to be scanned. The signal to noise ratio (SNR) cannot readily be increased by slower scanning or binning, because this would critically affect the temporal or spatial resolution required for particle tracking. Intravital 2PM at fast frame rates is a low light method in which, at least in dark image areas, only very few photons are collected per pixel. This unavoidably leads to a low SNR, which affects further data interpretation by human observers, and deteriorates the efficiency of automated processing, segmentation and image analysis [2,3,4]

Methods

Results

Conclusion