Adaptive Motion Artifact Reduction in Wearable ECG Measurements Using Impedance Pneumography Signal

Xiang An, Yanzhong Liu, Yixin Zhao, Sichao Lu, George K. Stylios, Qiang Liu2022 · DOI

TL;DR

This paper introduces a novel method for reducing motion artifacts in wearable ECG signals by using simultaneously measured Impedance Pneumography (IP) as a reference signal. By designing a specialized ECG-IP acquisition system and employing an adaptive filtering approach, they demonstrate a significant improvement in ECG signal quality, crucial for reliable remote health monitoring.

What Problem Does This Paper Address?

Wearable electrocardiogram (ECG) monitoring systems offer immense potential for continuous health tracking, early disease detection, and telemedicine. However, a persistent and significant challenge is the corruption of ECG signals by motion artifacts. These artifacts, arising from the wearer's movements and skin deformation, can mimic or mask actual cardiac events, leading to inaccurate interpretations, misdiagnoses, and ineffective treatment. Unlike other noise sources, motion artifacts are unpredictable and their spectral characteristics often overlap with vital ECG components, making their removal difficult. Existing methods, while showing promise, often struggle with the unpredictable nature of these artifacts or require external sensors like accelerometers, which can be cumbersome. This paper aims to address this by developing a more integrated and effective solution for motion artifact reduction in wearable ECGs.

Methods

The core of this research lies in an adaptive filtering technique combined with a novel data acquisition strategy.

  • Adaptive Filtering: The authors employ a Recursive Least Squares (RLS) adaptive filter. RLS is chosen for its faster convergence speed compared to other adaptive algorithms like LMS. To further refine the process, a 5 Hz low-pass filter is applied to the denoised ECG signal to mitigate high-frequency noise that can arise from aggressive filtering.
  • Impedance Pneumography (IP) as Reference Signal: The key innovation is the use of Impedance Pneumography (IP) as a reference signal for the adaptive filter. IP measures changes in thoracic electrical impedance, primarily due to respiration, but it also captures signals related to body movement and skin deformation. The hypothesis is that these motion-related components in the IP signal will correlate with the motion artifacts in the ECG.
  • ECG-IP Synchronous Acquisition System: To maximize the correlation between the IP signal and motion artifacts in the ECG, a custom hardware system was designed. This system uses the same pair of electrodes to simultaneously record both ECG and IP signals. This ensures that both signals are influenced by the same physical movements and environmental factors.
  • IP Signal Pre-processing: Before being fed into the adaptive filter, the IP signal is filtered with a 5 Hz low-pass FIR filter. This step is crucial to remove cardiac-related artifacts from the IP signal (which have a higher frequency than typical motion artifacts) while preserving the components that correlate with motion artifacts.

Datasets

The study utilizes a dataset collected from ten male and ten female healthy subjects. These subjects performed five specific upper-body motions designed to simulate common daily activities and trigger motion artifacts. Each subject repeated each motion ten times. This dataset was collected using the newly designed ECG-IP synchronous acquisition system. No pre-existing public datasets were used for evaluation; instead, the authors generated their own benchmark data under controlled experimental conditions.

Key Insights and Ideas

  • IP as a Viable Reference for Motion Artifacts: The most significant insight is that the Impedance Pneumography signal, when properly processed, can serve as an effective reference signal for adaptive motion artifact reduction in ECG. This is because the physical movements that cause motion artifacts in ECG also influence thoracic impedance.
  • Simultaneous Acquisition is Key: The design of a system that captures both ECG and IP signals using the same electrodes is critical. This ensures a high degree of correlation between the reference signal and the noise, which is a fundamental requirement for effective adaptive filtering.
  • Targeted Filtering of Reference Signal: Applying a specific low-pass filter (5 Hz) to the IP signal is essential. This selectively removes cardiac-related components from the IP signal, ensuring that the adaptive filter primarily learns to cancel the motion artifact components, not the actual cardiac signal.
  • Adaptive Filtering Outperforms Static Filtering: The results suggest that adaptive filtering, by its nature, can track and adapt to the changing characteristics of motion artifacts, offering a more robust solution than static filtering methods like a simple high-pass filter.

Key Results

  • Strong Correlation: The Pearson correlation coefficient between the 5 Hz low-pass filtered ECG and IP signals was found to be greater than 0.6 for most of the tested body motions, indicating a good correlation suitable for adaptive filtering.
  • Effective Artifact Reduction: The proposed adaptive motion artifact reduction method demonstrated good performance in reducing motion artifacts, as visually confirmed by the denoised ECG waveforms.
  • Quantitative Evaluation: The "goodness-of-fit" parameters, m_MSE and m_R², showed promising results. The m_MSE was consistently below 0.15, and m_R² was above 0.7, indicating that the estimated noise signal closely matched the actual motion artifact.
  • Superiority over IIR Filter: When compared to a standard 0.5 Hz IIR high-pass filter, the proposed adaptive method achieved a lower standard deviation (STD) in the denoised ECG signals, indicating more effective reduction of baseline drift and motion artifacts while preserving the ECG waveform.

Limitations and Future Work

  • Subject Population: The study was conducted on healthy subjects. The performance of the method on individuals with pathological ECG signals or more severe motion artifacts was not experimentally verified due to ethical considerations and the scope of the research.
  • Pathological ECG Preservation: While the authors hypothesize that their method would better preserve pathological ECG characteristics due to its frequency-domain approach, this was not directly tested.
  • Future Work: The authors suggest that their method could be applied to other vital sign monitoring systems that also suffer from motion artifact interference, such as respiration monitoring, especially in the context of the COVID-19 pandemic where wearable health systems are in high demand. Further exploration into real-world, unconstrained scenarios and diverse subject populations would be beneficial.

Summary updated

BibTeX
@misc{an2022adaptive,
  author = {Xiang An and Yanzhong Liu and Yixin Zhao and Sichao Lu and George K. Stylios and Qiang Liu},
  title = {{Adaptive Motion Artifact Reduction in Wearable ECG Measurements Using Impedance Pneumography Signal}},
  year = {2022},
  doi = {10.3390/s22155493},
}
Figure