Comprehensive evaluation and clinical implications of kernel extreme learning machine long short term memory...

Area of Science:

• Artificial Intelligence in Medicine
• Neuroscience
• Biomedical Informatics

Background:

• Alzheimer's disease (AD) diagnosis and prediction remain challenging.
• Accurate early risk stratification is crucial for timely intervention.
• Existing methods may not fully leverage complex clinical data.

Purpose of the Study:

• To develop and validate a hybrid deep learning model for enhanced AD diagnosis and prediction.
• To utilize readily available clinical data for improved accuracy.
• To assess the model's generalizability and predictive power.

Main Methods:

• A triple-architecture joint model integrating Kernel Extreme Learning Machine (KELM), Long Short-Term Memory (LSTM), and Transformer was developed.
• The model was trained and validated on 2,149 subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database.
• External validation was performed on an independent cohort of 1,012 subjects.

Main Results:

• Internal validation achieved 95.42% accuracy and 0.981 AUC.
• External validation demonstrated 93.81% accuracy and 0.9725 AUC.
• The model predicted mild cognitive impairment to AD conversion within 3 years with 92.78% accuracy.

Conclusions:

• The KELM-LSTM-Transformer model offers a powerful and robust framework for AD prediction.
• High accuracy and generalizability suggest its potential for early risk stratification.
• This accessible tool can support timely clinical interventions for Alzheimer's disease.