CINeMA: Conditional Implicit Neural Multi-Modal Atlas for a Spatio-Temporal Representation of the Perinatal Brain.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Medical Image Analysis

Background:

• Fetal and neonatal brain development involves rapid anatomical changes requiring high-resolution brain atlases.
• Existing atlas creation methods demand large datasets, posing challenges for studying rare pathologies.
• Accurate spatio-temporal brain models are crucial for understanding early neurodevelopment.

Purpose of the Study:

• To develop a novel framework, CINeMA (Conditional Implicit Neural Multi-Modal Atlas), for creating high-resolution, spatio-temporal, multimodal brain atlases.
• To address the limitations of data scarcity in studying developing brains, especially those with pathologies.
• To enable flexible atlas generation conditioned on various anatomical features and conditions.

Main Methods:

• CINeMA operates in latent space, bypassing computationally intensive image registration.
• The framework utilizes implicit neural representations for atlas construction.
• It allows conditioning on gestational age, birth age, and specific pathologies like agenesis of the corpus callosum and ventriculomegaly.

Main Results:

• CINeMA reduces atlas construction time from days to minutes.
• The framework achieves high accuracy, surpassing state-of-the-art methods.
• It supports downstream tasks like tissue segmentation and age prediction, and enables synthetic data generation.

Conclusions:

• CINeMA offers an efficient and versatile solution for creating advanced brain atlases, particularly in low-data scenarios.
• This framework facilitates research on early brain development and the impact of pathologies.
• The developed code and atlases are publicly released to advance the field.