Histogram-based features track Alzheimer's progression in brain MRI

Area of Science:

• Neuroscience
• Medical Imaging
• Computer Science

Background:

• Alzheimer's disease (AD) is a progressive dementia characterized by neurodegeneration, amyloid plaques, and neurofibrillary tangles.
• Early detection of AD is crucial for patient outcomes, with Magnetic Resonance Imaging (MRI) playing a key role in assessing neurodegeneration.
• Current computer-aided diagnosis tools face challenges in differentiating early dementia stages due to overlapping characteristics.

Purpose of the Study:

• To develop an adaptive multi-thresholding algorithm for identifying and tracking neurodegeneration progression in Alzheimer's disease.
• To define novel features for quantifying neurodegeneration stages from MRI data.
• To evaluate the efficacy of the proposed algorithm and features in classifying different stages of dementia.

Main Methods:

• An adaptive multi-thresholding algorithm based on smoothed histogram morphology was developed.
• Features such as gray/white matter volume, statistical moments, shrinkage, and geometric measures were mathematically derived.
• Multiple machine learning classifiers (Decision Tree, SVM, Naïve Bayes, KNN, Ensemble, Neural Network) were employed for evaluation.

Main Results:

• The proposed algorithm successfully defined features capable of identifying neurodegeneration.
• The derived features enabled tracking of disease progression through distinct stages: non, very mild, mild, and moderate.
• Experimental results demonstrated the successful labeling of neurodegeneration stages using the developed methodology.

Conclusions:

• The developed adaptive multi-thresholding algorithm and derived features show promise for accurate staging of neurodegeneration in Alzheimer's disease.
• This approach can aid clinicians in early diagnosis and monitoring disease progression.
• The methodology offers a quantitative tool for assessing the impact of Alzheimer's disease on the brain.