Peak-Detector: LLM-Based Explainable Peak Detection for Cardiac Signals
A research article available on arXiv (2605.16452) presents Peak-Detector, an innovative framework that leverages instruction-tuned Large Language Models (LLMs) to achieve reliable, cross-modal, and interpretable peak detection in cardiac physiological signals. This approach overcomes the shortcomings of traditional algorithms, which tend to be specific to certain modalities and lack clarity, as well as the opacity of deep learning techniques. Peak-Detector utilizes a 'peak-representation' method to convert time-series data into a more concise format, ensuring essential event details are retained. Targeting Electrocardiogram (ECG), Photoplethysmogram (PPG), Ballistocardiogram (BCG), and Bodyseismography (BSG) signals, the framework seeks to improve cardiovascular monitoring by enhancing both accuracy and transparency for expert validation.
Key facts
- Peak-Detector is a framework for peak detection in cardiac signals.
- It uses instruction-tuned Large Language Models (LLMs).
- The framework is explainable, addressing interpretability issues.
- It works across multiple modalities: ECG, PPG, BCG, BSG.
- A 'peak-representation' technique condenses time-series data.
- The paper is published on arXiv with ID 2605.16452.
- Conventional algorithms are limited to single signal modalities.
- Deep learning methods often lack interpretability.
Entities
Institutions
- arXiv