Identifying Leukoaraiosis with Mild Cognitive Impairment by Fusing Multiple MRI Morphological Metrics and Ensemble
Yifeng Yang1, Ying Hu2, Yang Chen1
1School of Health Science and Engineering, University of Shanghai for Science and Technology, 200093, Shanghai, People's Republic of China.
Journal of Imaging Informatics in Medicine
|February 12, 2024
Summary
An ensemble learning method using MRI features effectively identifies mild cognitive impairment in leukoaraiosis patients. This approach improves diagnostic accuracy for leukoaraiosis with mild cognitive impairment (LA-MCI) versus leukoaraiosis without cognitive impairment (LA-nCI).
Area of Science:
- Neurology
- Medical Imaging
- Machine Learning
Background:
- Leukoaraiosis (LA) is linked to cognitive decline and dementia risk.
- Accurate identification of LA patients with mild cognitive impairment (LA-MCI) is crucial for timely intervention and disease management.
Purpose of the Study:
- To develop and validate an ensemble learning model for distinguishing LA-MCI from LA without cognitive impairment (LA-nCI) using MRI-derived morphological features.
Main Methods:
- Extracted comprehensive MRI morphological features: gray matter volume, cortical thickness, surface area, cortical volume, sulcus depth, fractal dimension, and gyrification index.
- Employed an ensemble framework combining a data-level resampling method (Fusion+XGBoost) and an algorithm-level focal loss-improved XGBoost model (FL-XGBoost).
- Utilized the extreme gradient boosting (XGBoost) classifier within a weighted soft-voting ensemble to address class imbalance and enhance classification performance.
Main Results:
- The baseline XGBoost model achieved 78.20% balanced accuracy.
- Individual ensemble components improved performance: Fusion+XGBoost (80.53% Bacc) and FL-XGBoost (81.25% Bacc).
- The fused ensemble model achieved an overall accuracy of 84.82%, with sensitivity of 85.50% and specificity of 84.14% in distinguishing LA-MCI from LA-nCI.
Conclusions:
- The proposed ensemble learning method significantly enhances the accuracy and stability of differentiating LA-MCI from LA-nCI.
- This advanced MRI-based approach shows potential for facilitating clinical diagnosis and monitoring of cognitive impairment in LA patients.
