MRBrainS Challenge: Online Evaluation Framework for Brain Image Segmentation in 3T MRI Scans

Area of Science:

• Medical Imaging
• Neuroscience
• Computer Vision

Background:

• Numerous brain MRI segmentation methods exist, making algorithm selection challenging.
• Accurate segmentation of brain tissues is crucial for neurological research and clinical applications.

Purpose of the Study:

• To establish an online evaluation framework (MRBrainS) for assessing semi-automatic algorithms segmenting gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF).
• To provide an objective comparison of segmentation algorithms using standardized data and metrics.

Main Methods:

• Development of the MRBrainS online evaluation framework using 3T brain MRI scans from elderly subjects (65-80 years).
• Provision of training and testing datasets with manual segmentations as the reference standard.
• Evaluation of participating algorithms using Dice, H95, and AVD metrics, with results ranked and published online.

Main Results:

• Eleven segmentation algorithms and three freeware packages (FreeSurfer, FSL, SPM) were evaluated.
• The MRBrainS framework facilitated a direct and objective comparison of algorithm performance.
• Rankings based on overall performance for GM, WM, and CSF segmentation were established.

Conclusions:

• The MRBrainS evaluation framework offers a valuable tool for researchers to objectively compare and select the most effective brain MRI segmentation algorithms.
• This standardized evaluation process supports advancements in neurological research by improving the reliability of tissue segmentation.