Automated segmentation for cortical thickness of the medial perirhinal cortex

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Alzheimer's disease (AD) involves progressive neurofibrillary tangle (NFT) spread, starting in the medial perirhinal cortex (mPRC).
• The mPRC is a potential early diagnostic marker for AD due to its link with neuronal loss.
• Automated tools for measuring mPRC cortical thickness are limited.

Purpose of the Study:

• To develop and validate an automated tool for measuring cortical thickness in AD-vulnerable brain regions.
• To assess the utility of automated measurements for detecting structural changes in Alzheimer's dementia and mild cognitive impairment.

Main Methods:

• Utilized the nnU-Net framework to train models on structural MRI data from 126 adults.
• Applied trained models to an independent dataset of 103 adults (Alzheimer's dementia, amnestic mild cognitive impairment (aMCI), and healthy controls).
• Compared automated cortical thickness measurements with manual segmentations and analyzed group differences.

Main Results:

• Achieved high agreement between automated and manual cortical thickness measurements.
• Identified significant mPRC, entorhinal cortex (ERC), and lateral perirhinal cortex (lPRC) atrophy in Alzheimer's dementia compared to controls.
• Found significant ERC atrophy in the aMCI group compared to healthy controls.

Conclusions:

• The developed automated segmentation tool demonstrates high accuracy and agreement with manual measurements.
• The tool is effective in detecting regional brain atrophy associated with Alzheimer's disease and aMCI.
• This automated approach can significantly advance Alzheimer's disease research and early diagnosis.