Multi-template tensor-based morphometry: application to analysis of Alzheimer's disease.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Anatomy

Background:

• Tensor-based morphometry (TBM) is crucial for analyzing structural brain changes in neurological disorders.
• Conventional TBM relies on single-template registration, which can be susceptible to errors.
• Registration errors in TBM can impact the accuracy of disease classification and group comparisons.

Purpose of the Study:

• To introduce and evaluate novel multi-template TBM methods.
• To compare the performance of multi-template TBM against the conventional single-template approach.
• To enhance the accuracy and reliability of TBM in neurodegenerative disease research.

Main Methods:

• Developed and proposed four distinct multi-template TBM methodologies.
• Utilized a large dataset (N=772) from the ADNI database, including healthy controls, stable MCI, progressive MCI, and AD patients.
• Employed magnetic resonance (MR) imaging data for quantitative and qualitative performance evaluations.

Main Results:

• Multi-template TBM methods demonstrated statistically significant improvements over the single-template method.
• Achieved 86.0% accuracy in classifying controls versus AD subjects.
• Reached 72.1% accuracy in differentiating stable versus progressive MCI subjects.
• Generated smoother, more continuous group-level difference maps with higher t-values.

Conclusions:

• Multi-template TBM effectively compensates for registration errors, leading to enhanced classification performance.
• The proposed multi-template approaches offer a more robust and accurate analysis of brain morphology in neurodegenerative diseases.
• This method provides superior diagnostic imaging biomarkers for conditions like AD and MCI.