CLASSIC: consistent longitudinal alignment and segmentation for serial image computing.

Area of Science:

• Medical imaging
• Neuroscience
• Biomedical engineering

Background:

• Longitudinal magnetic resonance imaging (MRI) is crucial for tracking brain changes.
• Accurate segmentation of serial brain MR images is challenging due to anatomical variations and changes over time.
• Quantifying regional and global brain volume changes requires robust and consistent segmentation methods.

Purpose of the Study:

• To develop a temporally-consistent and spatially-adaptive algorithm for segmenting longitudinal brain MR images.
• To enable accurate measurement of rates of change in regional and global brain volumes.
• To estimate morphological changes like growth or atrophy from serial MR images.

Main Methods:

• The proposed algorithm, CLASSIC (Clustered Longitudinal Adaptive Spatially-adaptive Segmentation of Images of the Cranium), integrates image-adaptive clustering, spatiotemporal smoothness constraints, and image warping.
• It jointly segments a series of 3-D MR brain images from the same subject.
• The method accounts for changes due to development, aging, or disease.

Main Results:

• Experimental results on simulated and real longitudinal MR brain images demonstrate the algorithm's effectiveness.
• The CLASSIC algorithm achieves high segmentation accuracy.
• The method ensures longitudinal consistency in segmenting serial brain images.

Conclusions:

• The CLASSIC algorithm provides a robust solution for longitudinal brain MR image segmentation.
• It enables accurate quantification of brain volume changes and morphological alterations over time.
• This method is valuable for research in neurodevelopment, aging, and neurological diseases.