Project

Back to overview

Computational biomicroscopy: advanced image processing methods to quantify live biological systems

Applicant Liebling Michael
Number 179217
Funding scheme Project funding (Div. I-III)
Research institution IDIAP Institut de Recherche
Institution of higher education Idiap Research Institute - IDIAP
Main discipline Electrical Engineering
Start/End 01.04.2018 - 31.03.2022
Approved amount 630'749.00
Show all

All Disciplines (2)

Discipline
Electrical Engineering
Embryology, Developmental Biology

Keywords (7)

image processing; holography; computational photography; machine learning; microscopy; developmental biology; signal processing

Lay Summary (French)

Lead
La qualité des microscopes traditionnels repose sur d'excellents composants optiques. Avec la digitalisation des images, le calcul par ordinateur joue aujourd'hui un rôle prépondérant aussi bien pour la formation de l'image elle-même que pour l'analyse quantitative des objets imagés. La réduction du flou, l'amélioration de la capacité à distinguer des détails petits ou qui se déplacent rapidement en sont des exemples.
Lay summary

Le but de ce projet est de développer de nouveaux outils de calcul par ordinateur pour améliorer les microscopes. Ces outils s'appuieront sur des développements récents : dans le domaine de l'apprentissage machine (intelligence artificielle) et des représentations mathématiques parcimonieuses (traitement du signal). Nous visons, d'une part, à surmonter les limitations dues au manque de lumière et au flou en microscopie tridimensionnelle, en nous appuyant sur des techniques d'apprentissage pour déterminer des paramètres ouverts dans des modèles de déconvolution d'image et, d'autre part, à réduire des artéfacts dans la production et la reconstruction d'images de microscopie par holographie digitale, une technique qui permet de collecter des images avec de larges champs de vue et distances focales. Le développement de ces nouveaux outils de calcul devra permettre l'utilisation de composants optiques moins onéreux et de microscopes à construction simplifiée. 

 

Direct link to Lay Summary Last update: 18.04.2018

Responsible applicant and co-applicants

Employees

Associated projects

Number Title Start Funding scheme
164245 The Cellular Basis of Cardiac Development Revealead by Live Imaging 01.06.2016 Project funding (Div. I-III)
164245 The Cellular Basis of Cardiac Development Revealead by Live Imaging 01.06.2016 Project funding (Div. I-III)
164022 Platform for Reproducible Acquisition, Processing, and Sharing of Dynamic, Multi-Modal Data 01.07.2016 R'EQUIP
159227 Computational Methods for Temporal Super-resolution Microscopy 01.04.2016 Project funding (Div. I-III)

Abstract

Microscopes have traditionally relied on exquisite optical components. With the advent of image digitization, computation now plays an essential role, both in the imaging formation process as well as the quantitative analysis of the acquired images. The goal of this project is to develop imaging approaches that leverage two classes of computational tools: machine learning approaches and signal processing using sparse approximations. Their application in the general field of computational imaging is highly promising.Specifically, the first aim of the research in this project is to overcome resolution limitations in low-photon, three-dimensional fluorescence microscopy by use of machine learning techniques in deconvolution and optimal imaging parameter determination. The second aim is to address computational artifacts in digital holographic microscopy, a technique that allows collecting images over wide fields of view and focal ranges using fairly inexpensive hardware devices yet that requires carefully designed computational tools for post processing. There, we propose to work in a generalized sampling framework that can be adapted precisely to experimental acquisition devices while retaining the advantages of classical signal processing implementations (fast Fourier transforms). This framework is also particularly well-adapted for performing non-linear approximations in sparse signal representations. The proposed methods will be implemented and become part of an imaging platform recently set up at Idiap, which allows acquiring and disseminating microscopy data and tools in a reproducible research framework. The developed techniques should also allow future development of microscopes with inexpensive components and simplified hardware designs.
-