Automatic Segmentation of Hippocampus for Longitudinal Infant Brain MR Image Sequence by Spatial-Temporal Hypergraph...

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Developmental Neuroscience

Background:

• Accurate infant hippocampus segmentation is crucial for studying early brain development and neurological disorders.
• Manual segmentation is time-consuming and irreproducible.
• Existing automatic methods struggle with dynamic infant brain changes and temporal inconsistencies.

Purpose of the Study:

• To develop a novel method for accurate and consistent joint segmentation of infant hippocampi across longitudinal MRI scans.
• To address the challenges posed by dynamic anatomical changes in early life.
• To improve upon existing multi-atlas label fusion techniques.

Main Methods:

• A spatial-temporal hypergraph model is proposed to jointly segment infant hippocampi from longitudinal image sequences.
• The hypergraph encodes atlas-to-target relationships and spatial/temporal neighborhood information.
• A semi-supervised label propagation model is used for segmentation.

Main Results:

• The proposed method leverages spatial-temporal information for improved hippocampus segmentation.
• Experimental results show enhanced accuracy and consistency compared to state-of-the-art methods.
• The method was evaluated on T1-weighted brain MR images from 2 weeks to 12 months of age.

Conclusions:

• The spatial-temporal hypergraph model provides accurate and consistent infant hippocampus segmentation across multiple time-points.
• This approach is valuable for investigating early brain development and diagnosing neurological disorders.
• The method outperforms existing techniques by integrating longitudinal data effectively.