Automatic segmentation of amyloid plaques in MR images using unsupervised support vector machines

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Neuropathology

Background:

• Amyloid plaque deposition is a hallmark of Alzheimer's disease (AD).
• Accurate quantification of amyloid plaques in brain tissue is challenging.
• Existing methods lack the precision needed for detailed pathological analysis.

Purpose of the Study:

• To develop a novel, automatic algorithm for segmenting and quantifying β-amyloid (Aβ) plaques using MRI.
• To validate the algorithm's performance against histological data.
• To demonstrate the algorithm's utility in tracking age-related plaque changes in a mouse model of AD.

Main Methods:

• Utilized watershed transform to identify plaque candidates based on intrinsic MR signal characteristics.
• Employed unsupervised learning with MR intensity features for plaque classification.
• Validated results by comparing MRI-based segmentation with histology.
• Applied the algorithm to ex vivo MRI data from 5xFAD mice.

Main Results:

• Successfully developed an automatic plaque segmentation algorithm.
• Demonstrated high accuracy in plaque detection and quantification via comparison with histology.
• Showcased the algorithm's capability to detect age-dependent increases in plaque load in 5xFAD mice.
• Established a quantitative MRI method for characterizing amyloid plaques.

Conclusions:

• The novel algorithm provides a reliable, quantitative method for amyloid plaque characterization in MRI data.
• This technique facilitates the study of Aβ deposition spatiotemporal progression in AD mouse models.
• The method is valuable for evaluating the efficacy of novel amyloid-targeting therapies in preclinical settings.