Banca de QUALIFICAÇÃO: LUCAS CAIÃ DE SOUZA TAVARES

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
STUDENT : LUCAS CAIÃ DE SOUZA TAVARES
DATE: 10/04/2024
TIME: 09:30
LOCAL: Sala de Reuniões do ICe e https://meet.google.com/xeq-zskt-zhf
TITLE:

Machine Learning-Based Decoding of Task Variables Across the Mouse Brain


KEY WORDS:

brain-wide; decoding; dimensionality reduction; latent space; decision-making;


PAGES: 85
BIG AREA: Ciências Biológicas
AREA: Fisiologia
SUMMARY:

In this study, we explored the decoding of behavioral task variables from neuronal spiking data across 115 mice, spanning 267 brain regions with 547 Neuropixels insertions from 11 laboratories, provided by the International Brain Laboratory initiative. Mice were trained on a decision-making task, responding to visual stimuli with varying contrasts, influenced by probabilistic trial blocks. With this task setup, we mapped the embeddings of neural activity generated with the recent CEBRA dimensionality reduction algorithm and created a detailed brain-wide map of this latent space. Our work corroborates how specific brain regions contribute to the encoding of choice and feedback delivery, which display a broad spatial distribution, as well as visual stimuli and contrast levels, which are more localized. Utilizing a combination of the deep learning-based CEBRA and a KNN classifier, we demonstrate the capacity to decode these variables in different timeframes, highlighting that choice-related signals are most decipherable near movement initiation, whereas feedback signals maintain stable decodability across time. Notably, visual stimuli and contrast levels showed increased decodability some time after their presentation rather than immediately. These findings advance our understanding of the temporal and spatial dynamics of the neural activity underlying complex behaviors, offering new insights into the mechanisms of perception and decision-making in the brain.


COMMITTEE MEMBERS:
Presidente - 1721223 - ADRIANO BRETANHA LOPES TORT
Interno - 3086031 - DANIEL YASUMASA TAKAHASHI
Interno - 2069422 - DIEGO ANDRES LAPLAGNE

Notícia cadastrada em: 02/04/2024 17:13
SIGAA | Superintendência de Tecnologia da Informação - (84) 3342 2210 | Copyright © 2006-2024 - UFRN - sigaa04-producao.info.ufrn.br.sigaa04-producao