iBEAT V2.0: a multisite-applicable, deep learning-based pipeline for infant cerebral cortical surface reconstruction

Area of Science:

• Neuroimaging
• Developmental Neuroscience
• Computational Biology

Background:

• Infant brain MRI data present unique challenges due to low, dynamic tissue contrast and inter-site heterogeneity.
• Existing computational tools often perform poorly on infant brain magnetic resonance imaging (MRI) data, hindering accurate analysis of early brain development.
• Multisite infant neuroimaging studies require robust processing pipelines to handle variations in scanners and imaging protocols.

Purpose of the Study:

• To develop a robust, multisite-applicable computational pipeline specifically tailored for infant brain MRI analysis.
• To address the challenges of low tissue contrast and data heterogeneity in infant neuroimaging.
• To accurately quantify normal and abnormal early brain development using advanced deep learning techniques.

Main Methods:

• A novel deep learning-based computational pipeline was developed for infant brain MRI processing.
• The pipeline integrates preprocessing, brain skull stripping, tissue segmentation, topology correction, and cortical surface reconstruction/measurement.
• It handles both T1w and T2w structural infant brain MR images across a wide age range (birth to 6 years).

Main Results:

• The proposed pipeline demonstrates superior effectiveness, accuracy, and robustness compared to existing methods.
• It successfully processes diverse infant MRI datasets from multiple sites, scanners, and protocols, despite being trained on limited data.
• The iBEAT Cloud platform has processed over 16,000 infant MRI scans from more than 100 institutions.

Conclusions:

• The developed deep learning pipeline offers a robust solution for analyzing infant brain development from multisite MRI data.
• This tool effectively overcomes common challenges in infant neuroimaging, enabling more precise quantification of brain maturation.
• The widespread adoption and successful processing of numerous scans via iBEAT Cloud highlight the pipeline's practical utility and impact.