Deep Learning Algorithm Performs Automatic Segmentation of Neonatal Brains from MR Images

This imaging technique is particularly vital for neonates, especially for assessing neonatal encephalopathy, where it helps in understanding the presence and pattern of brain injuries for better prognostication and treatment planning. The integration of artificial intelligence (AI) and machine learning (ML) has significantly enhanced the predictive accuracy of functional outcomes in infants using MRI data. A crucial step in preparing data for ML analysis of brain MRI is brain extraction or skull-stripping. However, the development of extraction algorithms for neonatal brains has been limited. To address this gap, researchers have now introduced an automated deep learning-based algorithm for neonatal brain MRI extraction.

A collaborative effort by researchers from the University of California, San Francisco (UCSF) and Duke University Medical Center (Durham, NC, USA) has led to the creation of ANUBEX. This deep learning algorithm is specifically designed for automatic segmentation of neonatal brains from MRI scans. The development of ANUBEX, an automated neonatal nnU-Net brain MRI extractor, utilized various MRI sequences such as T1-weighted, T2-weighted, and diffusion-weighted imaging (DWI) from neonatal MRI studies.

The researchers found that ANUBEX maintains consistent performance when trained on sequence-agnostic or motion-degraded MRI scans, though it showed slightly decreased effectiveness on preterm brains. ANUBEX’s deep learning-based approach has demonstrated accurate performance across both high- and low-resolution MRIs, offering rapid computational processing. This accuracy in brain tissue segmentation is crucial for subsequent image analysis and volumetric measurements. Future directions for this research include expanding the evaluation of ANUBEX’s accuracy beyond the neonatal age range to include young children and adults. Additionally, there is a need to assess the model's effectiveness on brains with diverse structural pathologies.

Related Links:
UCSF
Duke University Medical Center