OPTIMAR consisted of five work packages (WPs).
WP1. Light propagation modelling | Mesoscopic imaging modalities work well for embryos and larvae that are highly transparent. However, for juvenile fish these methods are compromised due to scattering. These can cause stripes and shadows along the illumination direction that will affect the image quality and lead to incorrect quantitative conclusions. The Fellow, Dr. Teresa M Correia, developed accurate light propagation models (forward models) and angularly selective measurements of scattered and fluorescence light, in addition to transmitted light, which enabled quantitatively accurate reconstructions.
WP2. Image reconstruction from undersampled datasets | Imaging dynamic biological processes in 3D with high spatial and temporal resolution can be very challenging. Therefore, Dr. Correia developed image reconstruction strategies, based on the idea of compressed sensing, which generate high quality reconstructions from OPT and SPIM accelerated (undersampled) acquisitions, and thus, enable the visualisation of fast dynamic processes. However, compressed sensing-type algorithms do not achieve computational speeds compatible with real-time applications. Hence, a physics-informed deep learning image reconstruction method was proposed to accelerate scans and achieve real-time reconstructions. The proposed deep learning method successfully generates images from accelerated acquisitions in real time, whereas compressed sensing can take several minutes to reconstruct a complete 3D image.
WP3. Optimal source and detector sampling | The methods developed in WP2 were combined with illumination and detection strategies to: 1) reduce the number of projection images used for the OPT reconstruction; 2) reduce the number of z-stack images required for SPIM imaging. OPT was accelerated by acquiring less than 30 projections instead of several hundreds, whereas SPIM acquisitions were accelerated by 8-fold by using a patterned illuminations strategy.
WP4. Mesoscopic imaging system | An automated mesoscopic imaging system, controlled by computer, with 360-degree rotation, was set up at the host institute. The user-friendly software enables GPU-accelerated 3D reconstructions from 2D projection of transmitted and fluorescent using standard reconstruction methods and advanced deep learning-based reconstructions methods. In addition, it allows users to easily visualise and navigate 3D images.
WP5. Experimental studies using flatfish | Abnormal metamorphosis was induced by treatment with blockers of the thyroid axis. Control and treated flatfish larvae groups at different stages of development were studied using mesoscopic imaging and sequencing techniques. Flatfish sample preparation required testing and optimising tissue clearing and fluorescence labelling protocols. Our results indicate that the proposed mesoscopic imaging tools can be used to gain spatial and temporal insights into the hormone driven tissue and organ remodelling processes that occur during metamorphosis.
OPTIMAR outcomes have been shared on our websites (https://www.ccmar.ualg.pt/ |
https://qbioimaging.github.io/(s’ouvre dans une nouvelle fenêtre)) Twitter (@TeresaM2Correia | @CienciasDoMar), LinkedIn and other social media platforms.
