Computer aided progression detection model based on optimized deep LSTM ensemble model and the fusion of multivariate...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Informatics

Background:

• Alzheimer's disease (AD) is the leading cause of dementia, necessitating early and accurate detection for timely intervention.
• Current machine learning and deep learning (DL) models for AD detection show limited performance and stability in clinical settings.
• Ensemble learning methods often outperform standalone models, suggesting potential for improved AD prediction.

Purpose of the Study:

• To propose and evaluate a novel deep stacking framework for accurate early prediction of Alzheimer's disease.
• To leverage multivariate time-series patient data, including neuroimaging and cognitive scores, for enhanced AD detection.
• To develop a robust model for clinical decision support in managing AD progression.

Main Methods:

• Utilized a deep stacking ensemble framework combining multiple Long Short-Term Memory (LSTM) deep learning models.
• Employed Bayesian optimization to tune individual LSTM base classifiers on diverse feature sets.
• Trained heterogeneous base models on heterogeneous patient data to capture complex longitudinal patterns.
• Evaluated the ensemble model on a cohort of 685 patients from the National Alzheimer's Coordinating Center dataset.

Main Results:

• The proposed deep stacking ensemble model achieved high performance metrics: 82.02% accuracy, 82.25% precision, 82.02% recall, and 82.12% F1-score.
• Outperformed classical machine learning models and individual LSTM classifiers in predicting Alzheimer's disease.
• Demonstrated superior performance compared to existing state-of-the-art methods in the literature.

Conclusions:

• The novel deep stacking framework offers a significant advancement in early Alzheimer's disease detection.
• The ensemble model's high accuracy and stability make it a promising tool for clinical decision support.
• Further development could lead to a reliable system assisting experts in monitoring AD progression.