Benchmark on Automatic 6-month-old Infant Brain Segmentation Algorithms: The iSeg-2017 Challenge

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Developmental Neuroscience

Background:

• Accurate segmentation of infant brain magnetic resonance (MR) images into white matter (WM), gray matter (GM), and cerebrospinal fluid (CSF) is essential for understanding brain development and neurodevelopmental disorders.
• The isointense phase (6-9 months) presents segmentation challenges due to similar WM and GM intensities in T1w and T2w MR images, hindering accurate tissue differentiation.
• Few studies specifically address the segmentation of 6-month infant brain MR images, highlighting a gap in current research.

Purpose of the Study:

• To review and analyze the top-performing automatic segmentation methods from the iSeg-2017 challenge.
• To provide insights into the pipelines, implementations, and source codes of leading infant brain MRI segmentation techniques.
• To foster methodological development in the field of infant brain image analysis.

Main Methods:

• Review of the 8 top-ranked automatic segmentation methods from the iSeg-2017 challenge.
• Evaluation based on Dice ratio, modified Hausdorff distance, and average surface distance metrics.
• Analysis of the pipelines, implementations, and source codes of the reviewed methods.

Main Results:

• The review covers the 8 highest-ranked methods from the iSeg-2017 challenge, detailing their approaches to infant brain MRI segmentation.
• Performance metrics such as Dice ratio, modified Hausdorff distance, and average surface distance were used to rank the methods.
• Insights into the technical implementations and available source codes of these leading segmentation techniques are provided.

Conclusions:

• The iSeg-2017 dataset and this review offer valuable resources for advancing infant brain MRI segmentation methodologies.
• Understanding the strengths and limitations of current top methods can guide future research and development in this area.
• Further methodological development is needed to overcome the challenges in segmenting infant brains, particularly during the isointense phase.