Abstract
Long-term three-dimensional (3D) imaging of multicellular objects remains difficult due to strong multiple scattering, large spatio-temporal scales, and the need to preserve sample viability. Label-free approaches offer endogenous contrast but rarely combine high-throughput, temporal stability, and compatibility with thick, scattering specimens. We present here an automated and cost-effective framework integrating both hardware and software for time-lapse, phase-based imaging and analysis of 3D multicellular samples. The system enables parallel imaging of multiple specimens and integrates a complete acquisition-to-analysis pipeline enhanced by machine-learning algorithms. This approach minimizes manual intervention while improving robustness and reliability during multi-day experiments. Numerical reconstructions are performed using the beam propagation method as an iterative technique. The system was validated on hepatic organoids and mouse embryos, demonstrating high spatiotemporal resolution with limited phototoxicity. AI-based analysis focuses on the automated segmentation and tracking of selected structures of interest. Owing to its versatility and ease of implementation, our framework offers a practical solution for multi-samples, label-free time-lapse imaging and can be readily applied to different in vitro models.
Funding Information
  • European Commission (101016726)
  • Agence Nationale de la Recherche (ANR-21-CE19-0020)
  • French National Institute of Cancer (INCa_19463)
  • French National Institute of Cancer (N • 2024-167)

This publication has 47 references indexed in Scilit: