Classifying Alzheimer's Disease Using a Finite Basis Physics Neural Network.

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Alzheimer's disease (AD) is characterized by progressive neurodegeneration, leading to cognitive decline.
• Accurate classification of AD stages using functional magnetic resonance imaging (fMRI) faces challenges in data quality, interpretability, and standardization.
• Deep learning offers potential solutions for reliable AD classification from medical images.

Purpose of the Study:

• To propose a novel deep learning model, Classifying AD using a finite basis physics neural network (CAD-FBPINN), for accurate AD classification.
• To enhance fMRI image preprocessing and feature extraction for improved classification performance.
• To optimize the CAD-FBPINN model using the sea-horse optimization algorithm (SHOA) for superior AD staging.

Main Methods:

• Functional magnetic resonance imaging (fMRI) data were acquired from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.
• Images underwent preprocessing using a reverse lognormal Kalman filter (RLKF) and feature extraction via Newton-time-extracting wavelet transform (NTEWT).
• The extracted features were classified using a sea-horse optimization algorithm (SHOA)-optimized finite basis physics neural network (FBPINN).

Main Results:

• The proposed CAD-FBPINN method demonstrated significant improvements in accuracy, precision, and negative predictive value (NPV) compared to existing methods.
• Specifically, the method achieved higher accuracy (30.53%, 23.34%, 32.64%), precision (20.53%, 25.34%, 29.64%), and NPV (20.53%, 25.34%, 29.64%) over baseline approaches.
• The CAD-FBPINN technique outperformed other methods like DC-AD-AlexNet and PDP-ADI-GCNN in classifying various AD stages.

Conclusions:

• The CAD-FBPINN technique, optimized with SHOA, provides a robust and effective approach for classifying Alzheimer's disease stages using fMRI data.
• The method addresses key challenges in AD classification, offering potential for trustworthy and practical therapeutic applications.
• This deep learning approach shows promise for early detection and accurate staging of Alzheimer's disease.