Fig. 10: Comparative analysis of classification performance for MCI detection models using wearable EEG.
From: Wearable EEG devices in the detection of mild cognitive impairment: a systematic review
Radar charts comparing classification approaches across studies. Distance from center indicates performance (circles: 25%, 50%, 75%, 100%). Colors distinguish methodological approaches. A Unimodal versus multimodal features (8 studies, 2 rows): Studies systematically compared EEG-only features against multimodal integration incorporating physiological signals (HRV), behavioral measures (handwriting, eye-tracking), cognitive assessments, or multi-paradigm EEG. Multimodal fusion consistently achieved 5–15 percentage point improvements across studies. B Individual versus ensemble classifiers (12 implementations, 4 studies): Systematic comparison of individual machine learning algorithms (blue/cyan/green) against ensemble soft voting methods (red/orange). Ensemble approaches demonstrated consistent superiority with balanced performance across evaluation criteria. Multimodal integration and ensemble learning both substantially improve MCI classification, with combined approaches optimizing detection systems. MCI mild cognitive impairment, HRV heart rate variability, ERP event-related potential.
