En poursuivant votre navigation, vous acceptez l'utilisation de cookies destinés à améliorer la performance de ce site et à vous proposer des services et contenus personnalisés.


Deep learning tools and modeling to estimate the evolution of E2Fs concentration over the cell cycle from 2D still images

December 11, 2019 at 11 AM

Thierry Pecot, Medical University of South Carolina, USA

Lieu(x) :       Amphithéatre Lagarrigue, Ecole Polytechnique

Contact :    Anatole Chessel
                    anatole.chessel at polytechnique.edu

Since its appearance in 2012, deep learning has completely changed the conversation about artificial intelligence for image-based decision making. Deep learning tools developed over the last few years have outperformed all existing image segmentation approaches, from computer vision to medical imaging.

In this talk, three different deep convolutional neural networks are introduced and their performance compared for segmenting nuclei from complex tissues (mouse intestinal epithelium) in confocal and wide-field microscopy images. A similar approach is used to identify the markers associated with the nuclei. Once nuclei and their markers are well characterized, the intensity measured in the nuclei is used to estimate the temporal evolution of E2Fs, a family of transcription factors involved in cell cycle regulation, over the cell cycle. Finally, the correlation between the spatial distribution of E2Fs in the intestinal epithelium and their expression during cell cycle is demonstrated.