Continual Learning with Deep Neural Networks in Physiological Signal Data: A Survey

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Signal Processing

Background:

• Deep learning excels with physiological signals (ECG, EEG) but struggles with long-term monitoring's dynamic nature.
• Traditional models trained once lack adaptability to evolving healthcare data.
• Continual learning offers adaptive capabilities crucial for dynamic physiological signal analysis.

Purpose of the Study:

• To review continual learning techniques for physiological signal analysis.
• To explore applications and challenges of continual learning in smart healthcare.
• To bridge the literature gap on adaptive AI for long-term physiological monitoring.

Main Methods:

• Literature review of traditional and continual learning approaches.
• Analysis of continual learning techniques applied to ECG and EEG data.
• Discussion of implications for smart healthcare systems.

Main Results:

• Identified continual learning as a key solution for adaptive physiological signal processing.
• Outlined the evolution from static to adaptive deep learning models.
• Highlighted challenges including benchmarks, adaptability, efficiency, and user-centric design.

Conclusions:

• Continual learning is vital for advancing smart healthcare through adaptive physiological signal analysis.
• Future systems require focus on benchmarks, efficiency, and user needs.
• Further research is needed to establish robust continual learning frameworks for healthcare.