CogAdapt Framework Adapts Clinical ECG Models to Wearable Cognitive Load Assessment
A team of researchers has introduced CogAdapt, a framework designed to modify clinical ECG foundation models for assessing cognitive load in real time with wearable technology. This system features LeadBridge, an adaptable component that transforms 3-lead wearable signals into 12-lead formats, along with ProFine, a strategy for progressive fine-tuning that mitigates catastrophic forgetting. Tested on the CLARE and CL-Drive datasets using leave-one-subject-out cross-validation, CogAdapt surpasses traditional baseline approaches. This research tackles the issues of scarce labeled data and inadequate cross-subject generalization in the realm of adaptive human-computer interaction.
Key facts
- CogAdapt adapts clinical ECG foundation models to wearable cognitive load assessment
- LeadBridge converts 3-lead wearable signals into 12-lead representations
- ProFine is a progressive fine-tuning strategy that prevents catastrophic forgetting
- Evaluated on CLARE and CL-Drive datasets
- Leave-one-subject-out cross-validation was used
- CogAdapt substantially outperforms baseline methods
- Real-time cognitive load assessment is essential for adaptive human-computer interaction
- Clinical ECG foundation models are pre-trained on millions of recordings
Entities
—