Retracted: A Dynamic Adaptive Ensemble Learning Framework for Noninvasive Mild Cognitive Impairment Detection:...

Area of Science:

• Neuroscience
• Medical Technology
• Artificial Intelligence

Background:

• Early detection of mild cognitive impairment (MCI) is critical for preventing progression to severe neurodegenerative diseases.
• Current diagnostic methods are often costly, time-consuming, and invasive, limiting patient accessibility and compliance.
• There is a need for cost-effective, efficient, and noninvasive approaches to aid clinicians in MCI detection.

Purpose of the Study:

• To develop an ensemble learning framework for accurate and practical MCI detection.
• To integrate multimodal physiological data from wearable wristbands and digital cognitive metrics from tablets.
• To improve the accuracy and accessibility of early MCI diagnosis.

Main Methods:

• Recruited 843 participants aged 60+ for development and testing datasets.
• Collected physiological signals (electrodermal activity, photoplethysmography) and digital cognitive data.
• Employed a dynamic adaptive feature selection algorithm and optimized base learners for classification.

Main Results:

• Achieved classification accuracies of 88.4% (development), 85.5% (internal test), and 84.5% (external test).
• Area under the curve values ranged from 0.904 to 0.945 across datasets.
• Identified key indicators: skin conductance response decay time, heart rate variability (LF/HF ratio), and cognitive test completion time.

Conclusions:

• The developed MCI detection framework demonstrates high performance and stability in large-scale validation.
• Establishes a new benchmark for noninvasive, early MCI detection integrated into routine assessments.
• Enables convenient self-screening, mitigating healthcare access constraints and aiding the fight against neurodegenerative diseases.