A Novel Key Features Screening Method Based on Extreme Learning Machine for Alzheimer's Disease Study.

Area of Science:

• Computational Neuroscience
• Machine Learning in Medicine
• Biomedical Data Analysis

Background:

• Extreme Learning Machine (ELM) is an efficient algorithm for Single Hidden Layer Feedforward Neural Networks (SLFNs).
• ELM has been increasingly applied to Alzheimer's disease (AD) research, particularly for diagnosing AD using high-dimensional data.
• High-dimensional datasets often contain irrelevant or redundant features that can hinder diagnostic accuracy.

Purpose of the Study:

• To propose a novel Key Features Screening Method based on Extreme Learning Machine (KFS-ELM).
• To screen for and assign importance weights to key features relevant for Alzheimer's disease classification.
• To evaluate the effectiveness of KFS-ELM in improving AD diagnostic accuracy and understanding feature relevance.

Main Methods:

• Developed and applied the Key Features Screening Method based on Extreme Learning Machine (KFS-ELM).
• Experimentally screened 920 key functional connections from an initial set of 4005 functional connections for AD diagnosis.
• Assigned weights to the identified key features based on their contribution to classification.

Main Results:

• Diagnostic accuracy increased from 95.33% using all 4005 features to 99.20% using only the 920 key features identified by KFS-ELM.
• The 3085 features screened out negatively impacted AD diagnosis, confirming the effectiveness of KFS-ELM in feature selection.
• The KFS-ELM demonstrated a rational weighting system, where higher weights correlated with greater impact on AD diagnosis.

Conclusions:

• KFS-ELM is an effective method for screening key features and improving classification accuracy in Alzheimer's disease diagnosis.
• The method provides valuable insights into feature importance, aiding in the study of AD.
• KFS-ELM offers a robust tool for both feature analysis and enhancing diagnostic performance in complex medical datasets.