Host: The Japanese Society of Toxicology
Name : The 50th Annual Meeting of the Japanese Society of Toxicology
Date : June 19, 2023 - June 21, 2023
Drug-induced convulsive seizures are a serious adverse event and drug evaluation is usually carried out using behavioural indicators in laboratory animals such as rodents. However, these evaluation methods use large numbers of animals and require large amounts of test substance, resulting in low throughput. While tests using laboratory animals provide a lot of useful information, they are also considered to be an area for improvement in terms of animal welfare, speed of research and development and cost. To address this issue, a test method using zebrafish, which can be evaluated with a small amount of test substance (-µg) and can contribute to the promotion of the 3Rs, has been reported using the amount of behaviour immediately after drug addition as an indicator. However, in the previously reported evaluation system, the presence or absence of convulsions is finally determined by visual observation by researchers, which causes problems such as inconsistency in evaluation, reduced throughput and difficulty in multi-tracking. Therefore, in this study, we aimed to solve the above issues by introducing AI into the evaluation of seizure-like behaviour using zebrafish and automating some of the tasks.
In this presentation, an overview of automated tracking for each designated site using DeepLabCut, extraction of behavioural parameters that change in relation to compound administration, and evaluation of a seizure-like behaviour classification model using the behavioural parameter dataset will be presented.
Keywords: Toxicity prediction, Deep Learning, Multi-animal tracking, Central Nervous System
