Automatic hippocampal surface generation via 3D U-net and active shape modeling with hybrid particle swarm...

Area of Science:

• Neuroimaging
• Computational Anatomy
• Medical Image Analysis

Background:

• Hippocampal shape is a crucial biomarker for various neurological and psychiatric disorders.
• Accurate and automated hippocampal surface generation is essential for clinical research and diagnosis.
• Existing methods often require manual intervention or lack precision.

Purpose of the Study:

• To develop and validate a fully automatic pipeline for generating high-quality hippocampal surfaces.
• To integrate deep learning (3D U-net) with active shape modeling (ASM) for enhanced accuracy.
• To provide a reliable tool for biomarker extraction in brain disorder research.

Main Methods:

• A novel pipeline combining 3D U-net for segmentation and active shape modeling (ASM) with particle swarm optimization.
• Automatic hippocampus segmentation using 3D U-net on MRI data.
• Shape analysis via principal component analysis and optimization for surface generation.

Main Results:

• The pipeline successfully generated accurate hippocampal surfaces for both hemispheres.
• The generated surfaces exhibited correct anatomical topology and sufficient smoothness.
• High accuracy was achieved in matching segmented and generated surfaces.

Conclusions:

• The proposed automated pipeline offers a robust and efficient method for hippocampal surface generation.
• This tool has significant clinical relevance for the study of brain disorders.
• The integration of 3D U-net and ASM provides a powerful approach for neuroimaging analysis.