Related Experiment Video
Updated: Aug 1, 2025
05:56
Author Spotlight: Enhancing Diagnostic Strategies and Biomarker Development for Comprehensive Lung Function Analysis
Published on: August 9, 2024
1.6K
A Causal Intervention Scheme for Semantic Segmentation of Quasi-Periodic Cardiovascular Signals
IEEE Journal of Biomedical and Health Informatics
|April 27, 2023
Summary
Precise cardiac signal segmentation is improved by contrastive causal intervention (CCI). This method balances morphology and rhythm attributes, reducing bias for more objective deep learning representations and better anomaly detection.
Area of Science:
- Cardiovascular signal processing
- Deep learning for medical imaging
- Biomedical signal analysis
Background:
- Accurate cardiac cycle segmentation is crucial for analyzing cardiovascular signals and detecting anomalies.
- Deep semantic segmentation models often struggle with cardiovascular data due to over-reliance on morphology (A m) or rhythm (A r) attributes.
- Quasi-periodicity, a key characteristic of cardiovascular signals, is synthesized from both morphology and rhythm.
Purpose of the Study:
- To develop a novel deep learning approach that mitigates over-dependence on individual data attributes in cardiovascular signal segmentation.
- To introduce contrastive causal intervention (CCI) to create more objective and robust deep representations.
- To improve the accuracy of QRS location and heart sound segmentation.
Main Methods:
- Established a structural causal model to guide intervention strategies for morphology (A m) and rhythm (A r) attributes.
- Proposed contrastive causal intervention (CCI) within a frame-level contrastive learning framework.
- Conducted experiments under controlled conditions for QRS location and heart sound segmentation, utilizing multiple databases and noisy signals.
Main Results:
- The proposed contrastive causal intervention (CCI) approach significantly improved segmentation performance.
- Performance gains of up to 0.41% for QRS location and 2.73% for heart sound segmentation were observed.
- The method demonstrated generalizability across different databases and robustness to noisy cardiovascular signals.
Conclusions:
- Contrastive causal intervention (CCI) effectively addresses the bias issue in deep semantic segmentation of cardiovascular signals.
- The approach leads to more objective representations by balancing morphological and rhythm attributes.
- CCI offers a promising training paradigm for enhancing the analysis of quasi-periodic biomedical signals.
