Real-time cell population analysis in bioreactors using deep learning-enabled in situ microscopy - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2024

Real-time cell population analysis in bioreactors using deep learning-enabled in situ microscopy

Pierre-Yves Lacroix
Valdinei Luis Belini
  • Fonction : Auteur
  • PersonId : 1402942
Hajo Suhr
  • Fonction : Auteur
Philipp Wiedemann
  • Fonction : Auteur
  • PersonId : 932204
Thierry Chateau

Résumé

During a biotechnological production process using living cells, a substantial number of cells can enter physiological routes which are undesirable, directly lowering the productivity of the process or/and the product quality, whatever the product of interest is. When these cell cultures are performed in bioreactors, there are several critical key parameters of the process that can be monitored to better control the process and prevent cells from deviations. Indeed, cells are subjected to different types of environmental stress, such as hydrodynamic shearing or concentration gradients, exposing them to varying external conditions of dissolved oxygen and carbon dioxide, pH, nutrients and by-products that affect their physiology, which can be amplified when the operation is carried out in large-scale bioreactors (Paul and Herwig, 2020). It must be pointed out that the parameters involving cells represent mean values that do not express the behavior and the heterogeneity at the individual cell level. Extrinsic heterogeneity triggered by process parameters should also be completed with the insights obtained on the intrinsic heterogeneity caused by the stochasticity of both gene expression and metabolic reactions, at single-cell level (Delvigne et al., 2014). As heterogeneity globally affects the performance, it raises the importance of developing new methods for bioprocess development and monitoring, which correlate cell cultivation and cell-to-cell analysis, preferably in real-time (Schmitz et al., 2019). According to the regulatory boundaries set by the European EMA or the U.S. FDA, monitoring is also necessary to document repeatability as quality control and ensure a proper operation regime. For these purposes, numerous process analytical tools exist (Rathore et al., 2021). But the state-of-the-art shows that cell population analysis techniques do not provide real-time information with respect to morphological heterogeneity of cells during cultivation in bioreactors, neither at the industrial scale nor for R&D purposes. Here we report the development of an artificial intelligence-based tool applied to high-resolution online in situ microscopy images, aiming to provide accurate cell population analysis according to well-known cellular morphological features within biopharmaceuticals and bioenergy processes.
Fichier non déposé

Dates et versions

hal-04652252 , version 1 (18-07-2024)

Identifiants

  • HAL Id : hal-04652252 , version 1

Citer

Jean-Sébastien Guez, Pierre-Yves Lacroix, Valdinei Luis Belini, Hajo Suhr, Philipp Wiedemann, et al.. Real-time cell population analysis in bioreactors using deep learning-enabled in situ microscopy. 14th European Symposium on Biochemical Engineering Science ESBES 2024, Oct 2024, Copenhagen, Denmark. ⟨hal-04652252⟩
0 Consultations
0 Téléchargements

Partager

Gmail Mastodon Facebook X LinkedIn More