Deep ensemble learning of sparse regression models for brain disease diagnosis

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging Analysis

Background:

• Machine learning, particularly sparse regression, is effective for high-dimensional medical data with limited samples.
• Deep learning methods demonstrate state-of-the-art performance across various applications.
• Accurate diagnosis and prognosis of Alzheimer's disease and mild cognitive impairment are critical.

Purpose of the Study:

• To propose a novel framework combining sparse regression and deep learning for Alzheimer's disease/mild cognitive impairment diagnosis and prognosis.
• To develop a Deep Ensemble Sparse Regression Network (DESRN) for enhanced clinical decision-making.

Main Methods:

• Multiple sparse regression models were trained with varying regularization parameters to select diverse feature subsets.
• Response values from sparse regression models were used as target-level representations.
• A deep convolutional neural network was built upon these representations to form the DESRN.

Main Results:

• The proposed DESRN achieved the highest diagnostic accuracies in three classification tasks on the ADNI cohort.
• The method demonstrated effectiveness in handling high-dimensional brain imaging data for disease diagnosis.
• Rigorous analysis and comparison with existing literature validated the findings.

Conclusions:

• The integration of sparse regression and deep learning offers a powerful approach for Alzheimer's disease and mild cognitive impairment diagnosis.
• The DESRN framework represents a significant advancement in computer-assisted intervention for neurodegenerative diseases.
• This study highlights the potential of novel machine learning combinations for improving clinical outcomes.