Computational SPIM with AI Reconstruction Enhancements

Introduction

In recent years, the field of computational microscopy has witnessed remarkable advancements, transforming the way we acquire and analyze three-dimensional (3D) biological samples. Among these innovations, the Selective Plane Illumination Microscopy (SPIM) technique has emerged as a powerful tool for non-invasive 3D imaging of biological specimens. Coupled with artificial intelligence (AI) reconstruction techniques, SPIM microscopy has the potential to improve our understanding of complex biological structures. This training subject aims to explore the integration of SPIM microscopy with AI-based image reconstruction, focusing on image registration using Convolutional Neural Networks (CNN).

The Research Laboratory

This training subject is offered by CREATIS laboratory. Located at Université Claude Bernard Lyon1/INSA, our laboratory boasts state-of-the-art facilities and a multidisciplinary team of researchers dedicated to advancing the frontiers of biomedical imaging.

Scientific Context

The training will immerse participants in the dynamic intersection of computational microscopy, AI, and image registration. Participants will explore the following key areas:

  1. SPIM Microscopy Fundamentals: An in-depth understanding of SPIM microscopy principles, including light-sheet illumination, sample preparation, and data acquisition techniques. Participants will gain hands-on experience with SPIM setups.
  2. Artificial Intelligence in Microscopy: Exploring the integration of AI, particularly CNNs, for image reconstruction and enhancement. Participants will learn to leverage neural networks to improve image quality, denoise data, and extract valuable structural information.
  3. Image Registration: Detailed exploration of image registration techniques, with a focus on aligning and stitching multiple SPIM datasets. Participants will develop expertise in geometric and intensity-based registration methods, optimizing data fusion.

Outcome

Develop techniques for fusing the spectral and spatial information from HyperSpectral-SPIM (HS-SPIM) and Standard SPIM (S-SPIM) datasets. This should result in a single, integrated hypercube that combines the benefits of both detection arms for comprehensive analysis. 

  1. Image Registration Framework: Develop a robust and efficient image registration framework that can align HS-SPIM and S-SPIM images accurately.
  2.  Synergistic Reconstruction: Develop a technique to collaboratively reconstruct a hypercube using both the HS-SPIM and S-SPIM datasets, aiming to achieve a hypercube with combined spectral resolution from HS-SPIM and spatial resolution from S-SPIM.

Upon completing this training, participants will possess advanced skills in computational SPIM microscopy and AI-enhanced image reconstruction. They will be equipped to apply these techniques to their own research projects, contributing to the advancement of biological sciences. Join us at CREATIS to embark on this exciting journey at the cutting edge of computational microscopy. Together, we aim to unlock new insights into the mysteries of life through the synergy of SPIM microscopy and artificial intelligence.

 

For further information or inquiries about this training subject, please contact: 

Cédric Ray Garreau <cedric.ray@univ-lyon1.fr> or Nicolas Ducros <Nicolas.Ducros@insa-lyon.fr>