Multi-atlas label fusion with random local binary pattern features: Application to hippocampus segmentation.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Machine Learning

Background:

• Accurate hippocampus segmentation in MRI is crucial for neuroimaging studies.
• Existing methods may lack the precision needed for detecting subtle volumetric changes.

Purpose of the Study:

• To develop and validate a novel feature extraction method for improved hippocampus segmentation.
• To assess the performance of the proposed method in identifying volumetric differences associated with Alzheimer's disease and mild cognitive impairment.

Main Methods:

• Implemented a multi-atlas image segmentation (MAIS) framework using a novel random local binary pattern (RLBP) feature extraction.
• Registered selected atlases to target MR images using non-linear registration.
• Trained linear regression models based on RLBP features for segmentation.

Main Results:

• The RLBP-based MAIS method demonstrated competitive accuracy compared to state-of-the-art label fusion techniques.
• The method effectively detected volumetric differences in the hippocampus between Alzheimer's disease patients, mild cognitive impairment subjects, and normal controls.
• The algorithm showed good performance in segmenting hippocampus from 135 T1 MR images from the ADNI database.

Conclusions:

• The RLBP-based MAIS method offers efficient and accurate hippocampus segmentation.
• This technique shows promise for aiding in the prediction and diagnosis of Alzheimer's disease.
• The developed algorithm facilitates the analysis of hippocampal volume changes in neurological disorders.