Consistent Multi-Atlas Hippocampus Segmentation for Longitudinal MR Brain Images with Temporal Sparse Representation.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Accurate segmentation of the hippocampus is crucial for studying neurodegenerative diseases.
• Longitudinal studies require consistent and precise segmentation across multiple time points.
• Existing methods often struggle with temporal consistency in longitudinal hippocampus segmentation.

Purpose of the Study:

• To develop a novel multi-atlas based longitudinal label fusion method for simultaneous hippocampus segmentation.
• To improve the accuracy and temporal consistency of hippocampus segmentation in longitudinal imaging data.
• To leverage temporal sparse representation for robust label propagation.

Main Methods:

• Groupwise longitudinal registration to create a consistent subject image sequence.
• Longitudinal atlas alignment to the subject's group-mean image.
• A novel longitudinal label fusion technique incorporating temporal sparse representation for voxel labeling.

Main Results:

• The proposed method achieves simultaneous segmentation of hippocampi across all time points.
• Experimental results show superior accuracy and consistency compared to state-of-the-art methods.
• Demonstrated robustness in propagating atlas labels with temporal constraints.

Conclusions:

• The developed method offers a significant advancement in longitudinal hippocampus segmentation.
• It provides a more accurate and temporally consistent approach for analyzing brain changes over time.
• This technique has potential applications in clinical research and neurodegenerative disease monitoring.