Ressource pédagogique : Dr Natalie Reznikov - Application of deep learning for segmentation of 3D images in biomineralization

Mots-clés :
cours / présentation - Date de création : 23-03-2021
Partagez !

Présentation de: Dr Natalie Reznikov - Application of deep learning for segmentation of 3D images in biomineralization

Informations pratiques sur cette ressource

Langue du document : Anglais
Type pédagogique : cours / présentation
Niveau : enseignement supérieur
Durée d'exécution : 20 minutes 46 secondes
Contenu : image en mouvement
Document : video/mp4
Taille : 215.32 Mo
Droits d'auteur : libre de droits, gratuit
Droits réservés à l'éditeur et aux auteurs. @ LE STUDIUM 2021

Description de la ressource pédagogique

Description (résumé)

Modern 3D imaging methods in biomineralization – such as X-ray tomography and dual-beam electron tomography – produce datasets that are rich in fine detail and enormous in size, often containing inevitable artifacts.  Rendering segmentations of such datasets is a daunting task.  The recent introduction of artificial neural network-based deep learning into bioimaging has made 3D segmentation reliable, accurate and fast.  A highlight of convolutional neural networks (CNNs) is that artificial "neurons" are interlinked hierarchically, similarly to how feature-forming patterns of an image are related.  Accordingly, when a raw image is presented to a deep net, the neurons of different layers perceive the patterns of different complexity.  Upper-level neurons detect small patterns within their local context, and the local context itself forms patterns for deeper neuronal layers, and within a larger context, and so on.  Thus, identification of features based on overt (e.g. contrast, gradient) and covert patterns (e.g. level of noise, wavelet frequency) becomes not only accurate, but also generalizable.  Once image patterns can be accurately enough identified as being features of interest – and thus the CNN is “trained” – such patterns can be segmented automatically on any similar image.  In machine learning, as in biological learning, the accuracy of pattern detection and classification improves with experience.  Once trained, a CNN can be treated like an image filter – easy to preview, fast to apply, simple to share, and handy to reuse.  In this presentation, I will explain the essence of deep learning and CNN operation for non-computer scientists, and will illustrate this with examples of “difficult” 3D images (a chick embryo inside a fertilized egg, and coral).

"Domaine(s)" et indice(s) Dewey

  • Biochimie (572)

Thème(s)

Document(s) annexe(s) - Dr Natalie Reznikov - Application of deep learning for segmentation of 3D images in biomineralization

Partagez !

EN SAVOIR PLUS

  • Identifiant de la fiche
    60603
  • Identifiant
    oai:canal-u.fr:60603
  • Schéma de la métadonnée
  • Entrepôt d'origine
    Canal-U