Microscopic image analysis for the automation of toxicity and efficacy studies through zebrafish model. 3D Synthetic Modelling

The aim of the project is to digitise and automate the processes of pharmacological tests on zebrafish larvae at Biobide, in order to save the time spent by the technicians on zebrafish larvae at Biobide, in order to save the time spent by laboratory technicians in these processes, while increasing intra- and inter-observer laboratory technicians in these processes, while increasing intra- and inter-observer concordance.
All this, using Artificial Intelligence (e.g. Deep Learning) applied to the analysis of images of zebrafish larvae; and analysis of zebrafish larvae; and automated generation of synthetic databases with fine granularity annotations for the granularity annotations, for the development of highly accurate deep learning models. This project focuses on This project focuses on automating toxicity and cardiopathy (cardio) studies.
Vicomtech has the role of an expert in artificial intelligence and 3D modelling and is addressing three main challenges:

• Identification of appropriate pose and state of the larva during the image acquisition process using image processing and Deep using image processing and Deep Learning.
• Detection of endpoints (diseases or deformations) of the larva that indicate adverse drug effect using Deep Learning.
• Generation of annotated synthetic images (i.e. region occupying, landmarks, endpoints) from 3D modelling and rendering of images, using Blender®. This This last point is key and innovative due to the difficulty of obtaining enough annotated images and with the level of detail annotated and with the required level of detail. This approach adds the additional challenge of making the images as similar as possible to reality. For the latter, domain adaptation domain adaptation" Deep Learning techniques.

BIOBIDE, the project coordinator, is a benchmark in the field of toxic substance screening and its efficiency in a large catalogue of models. its efficiency in a large catalogue of disease models that have been developed over more than 16 years in the field. more than 16 years of experience in the market. Among its main clients are the NIEH in the USA and the main multinational companies in the field of cosmetics, signature and agrochemicals.
agrochemicals.
The Daniobit project has been funded by the HAZITEK 2022. Programme for the support of R&D Business, within the Research and Development Projects of a competitive nature, with the file number: ZL-2023/00220