Biomarker extraction-based Alzheimer's disease stage detection using optimized deep learning approach.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Alzheimer's disease (AD) is characterized by progressive cognitive decline and memory loss.
• Early diagnosis of AD is critical for effective treatment and preserving daily function.
• Current AD detection methods face challenges including model overfitting, underutilized biomarkers, and noisy imaging data.

Purpose of the Study:

• To develop a novel deep learning framework for improved identification of Alzheimer's disease stages.
• To address limitations in existing AD diagnostic techniques, such as overfitting and data noise.

Main Methods:

• Utilized structural MRI scans as the primary diagnostic data.
• Applied image enhancement techniques (histogram equalization, wavelet soft thresholding) for noise reduction.
• Implemented biomarker segmentation (ventricular, hippocampal) optimized with a firefly algorithm.
• Performed dimensionality reduction using Linear Discriminant Analysis to mitigate overfitting.
• Employed a Deep Belief Network optimized via Cuckoo Search for classification and feature learning.

Main Results:

• The proposed deep learning framework demonstrated enhanced accuracy in AD stage detection.
• Achieved a 0.66% increase in accuracy and a 0.0345% decrease in error rate compared to existing methods.

Conclusions:

• The developed deep learning strategy shows significant promise for early and accurate Alzheimer's disease stage identification.
• Improvements in segmentation, dimensionality reduction, and classification contribute to the framework's enhanced diagnostic performance.
• This approach represents a meaningful advancement in the field of Alzheimer's disease diagnostics.