Alzheimer disease detection from structural MR images using FCM based weighted probabilistic neural network

Area of Science:

• Neuroscience and Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Early intervention for Alzheimer's disease (AD) is crucial for patient outcomes and treatment efficacy.
• Accurate classification of Normal Control (NC), Mild Cognitive Impairment (MCI), and AD is essential for timely management.

Purpose of the Study:

• To develop and validate a novel classification algorithm for discriminating between NC, MCI, and AD using structural MRI.
• To enhance classification performance by integrating supervised and unsupervised learning techniques and removing noisy training data.

Main Methods:

• A novel Fuzzy C-Means based Weighted Probabilistic Neural Network (FWPNN) classification algorithm was developed.
• Structural MRI brain images were normalized, and regions of interest (Hippocampus and Posterior Cingulate Cortex) were extracted using Automated Anatomical Labeling (AAL).
• Nineteen relevant features were selected using a Multiple-criterion feature selection method, and noisy samples were removed from training data.

Main Results:

• The FWPNN algorithm achieved high classification accuracies: 98.63% (AD vs. NC), 95.4% (MCI vs. NC), and 96.4% (AD vs. MCI).
• Validation was performed on the ADNI subset and the Bordex-3 city dataset.
• Removal of noisy samples from training data significantly improved the decision-making process of the expert system.

Conclusions:

• The proposed FWPNN classification approach demonstrates high reliability and accuracy in discriminating between Alzheimer's disease stages using structural MRI.
• Integrating supervised and unsupervised learning with noise reduction techniques offers a promising strategy for improving diagnostic accuracy in neurodegenerative disease classification.