Accurate and Fully Automatic Hippocampus Segmentation Using Subject-Specific 3D Optimal Local Maps Into a Hybrid...

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Hippocampus (HC) structural integrity is crucial for understanding and diagnosing brain disorders.
• Accurate, reliable, and reproducible automatic HC segmentation methods are in high demand.
• Existing methods often require manual parameter tuning, limiting efficiency.

Purpose of the Study:

• To present an innovative 3D fully automatic method for hippocampus segmentation.
• To eliminate the need for heuristic parameter fine-tuning in segmentation.
• To enhance the anatomical suitability of segmentation for test images.

Main Methods:

• Developed a 3D fully automatic method utilizing a multiatlas concept.
• Employed subject-specific 3D optimal local maps (OLMs) to control active contour model (ACM) energy terms.
• Defined OLMs and optimal ACM parameters simultaneously via an optimization scheme, avoiding manual tuning.

Main Results:

• The proposed method demonstrated high accuracy and robustness across three public datasets.
• Achieved comparable or superior performance against state-of-the-art segmentation techniques.
• Successfully produced anatomically suitable OLMs for test images through an extended multiatlas concept.

Conclusions:

• The novel 3D automatic method offers an accurate, reliable, and reproducible solution for hippocampus segmentation.
• Simultaneous optimization of OLMs and ACM parameters eliminates the need for manual fine-tuning.
• This approach holds significant potential for the prevention, diagnosis, and follow-up of brain disorders.