Introduction to Photoplethysmography and Its Challenges
Smartwatches and other wearable devices are equipped with photoplethysmography (PPG) sensors for monitoring heart rate and other aspects of cardiovascular health. However, PPG signals collected from such devices are susceptible to corruption from noise and motion artifacts, resulting in inaccuracies during heart rate estimation. Conventional denoising methods filter or reconstruct signals in ways that eliminate morphological information, even from the clean segments of the signal that should ideally be preserved.
In this work, we develop an algorithm called SPEAR (Self-supervised PPG Erase Artifacts and Reconstruct) for denoising PPG signals. SPEAR leverages a large database of clean PPG signals to train a denoising autoencoder in a self-supervised manner. The algorithm identifies noisy regions in a signal, removes them, and reconstructs the corrupted regions to produce a clean signal. As we demonstrate, the reconstructed signals provide better estimates of heart rate from PPG signals than leading heart rate estimation methods. Further experiments show improvement in heart rate variability (HRV) estimation from PPG signals using our SPEAR algorithm.
Photoplethysmography: Principles and Applications
Photoplethysmography (PPG) is a noninvasive optical measurement technique that provides vital information about the cardiovascular system. A PPG-enabled device consists of an optical sensor that measures volumetric variations of blood circulation as a PPG signal. Modern PPG-enabled devices include a variety of technologies such as fingertip-based pulse oximeters, forehead- and earlobe-based PPG sensors, and most commonly, wrist-worn smartwatches.
PPG monitoring can enable early detection of serious heart conditions that otherwise might go undetected. A key application of PPG in wearable devices is the estimation of heart rate (HR). However, PPG signals are limited by their susceptibility to noise artifacts, including motion artifacts (MA) caused by body movements and artifacts arising from environmental factors like ambient light, sweat, and pressure.
Limitations of Conventional Denoising Methods
Methods that address the artifact problem for prediction of HR from PPG signals can be broadly categorized into two types. The first type estimates HR directly from the signals despite the presence of artifacts. The second type attempts to extract, denoise, or reconstruct a clean signal from the noise-corrupted signal.
These signal reconstruction approaches have some limitations. They reconstruct the entire PPG signal, even if most of the signal may already be artifact-free. This may potentially distort the original signal and cause loss of morphological information even in the useful parts of the signal. Ideally, we would like to have a method that denoises only the noisy part of the signal, preserving the valuable information in the uncorrupted part, and provides a clean signal that can be used for accurate HR estimation and for other downstream tasks.
The SPEAR Algorithm
SPEAR is a novel algorithm for denoising PPG signals that reconstructs the corrupted parts of the signal, while preserving the clean parts of the PPG signal. Our framework relies on self-supervised training, where we leverage a large database of clean PPG signals to train a denoising autoencoder.
The key steps in the SPEAR algorithm are:
- Removal of Artifacts: An artifact-detection algorithm is used to remove segments with artifacts from the PPG signal, leaving only the clean signals.
- Erasing Random Parts: Random parts of the clean signals are erased, creating partially corrupted samples.
- Training the Denoising Autoencoder: A denoising autoencoder is trained to reconstruct the erased parts of the clean signals.
During testing, given a new noisy signal, SPEAR would (1) apply the artifact-detector, (2) erase the artifacts, and (3) reconstruct the missing pieces using the trained denoising autoencoder to form a clean signal that can be used for downstream tasks.
Since the denoising autoencoder was trained to reconstruct from clean PPG signals, it will output clean signals during testing. This approach preserves the valuable information in the uncorrupted parts of the signal, while only reconstructing the corrupted regions.
Experimental Evaluation
We evaluated the performance of SPEAR on two publicly available PPG data sets: the Stanford data set and the PPG-DaLiA data set. The Stanford data set is a large collection of PPG signals from wrist-worn wearables, while the PPG-DaLiA data set contains PPG recordings along with synchronous electrocardiogram (ECG) signals, which provide ground-truth heart rate measurements.
We compared SPEAR against several state-of-the-art baselines, including signal processing techniques and deep learning-based methods for both heart rate estimation and heart rate variability (HRV) estimation.
Heart Rate Estimation
The experimental results reveal that traditional signal processing techniques generally achieve limited efficacy in heart rate estimation, and that supervised deep learning methods show better estimation accuracy on data sets they are trained on, with diminished generalizability to other data sets. SPEAR, on the other hand, does not exhibit these limitations.
On the PPG-DaLiA test set, SPEAR’s performance is comparable to that of deep learning methods trained on the same distribution, despite SPEAR being trained on the Stanford data set. On a hold-out test set from the Stanford data set, SPEAR outperforms all other methods.
The fact that SPEAR produces clean, continuous PPG signals allows the results to be used for downstream tasks beyond heart rate estimation. This is in contrast with other approaches that work on small, overlapping segments of the signal.
Heart Rate Variability Estimation
Heart rate variability (HRV) measures the fluctuation in the time intervals between adjacent heartbeats and is a valuable metric for assessing cardiac health. Studies that use PPG signals to estimate HRV have focused only on clean signals obtained from subjects at rest, showing poor performance under free-living conditions.
We found that the denoised signals produced by SPEAR enhanced the accuracy of HRV estimation, surpassing the performance of other methodologies. SPEAR achieved an improvement of approximately 62% over the original signal and 34% over a baseline denoising method in estimating the standard deviation of inter-beat intervals (SDNN). For the root mean square of successive differences (RMSSD), SPEAR achieved an improvement of approximately 63% over the original signal and 39% over the baseline.
These results demonstrate that SPEAR’s denoising algorithm yields significant improvements on the task of HRV estimation from PPG signals, enabling more reliable health monitoring in real-world conditions.
Practical Applications and Integration
Photoplethysmography (PPG) technology is becoming increasingly ubiquitous with the adoption of modern wearables such as smartwatches, wristbands, and smart jewelry. These devices allow users to continuously monitor heart rate and other health metrics throughout daily life.
PPG has several personal health applications, such as tracking blood pressure, blood oxygen saturation, monitoring sleep quality, and guiding exercise and recovery. Continuous long-term monitoring of PPG also has important clinical applications, such as diagnosing cardiovascular diseases and arrhythmia.
However, PPG signals are susceptible to noise and motion artifacts, limiting the accuracy and reliability of these applications. SPEAR can be integrated as a preprocessing step for any application that uses PPG recordings for predictive or analytical tasks. The reconstructed clean signals produced by SPEAR can result in more reliable performance in downstream tasks.
SPEAR’s code is publicly available and open source, allowing it to be easily integrated into personal health applications as well as clinical data-processing pipelines. This can enable the development of a variety of PPG-enabled applications available to the public, without being limited to the proprietary algorithms provided by device manufacturers.
Conclusion
In this work, we presented SPEAR, a novel self-supervised algorithm for denoising photoplethysmography (PPG) signals. SPEAR outperforms state-of-the-art methods in heart rate estimation and heart rate variability estimation, while preserving the clean parts of the signal and only reconstructing the corrupted regions.
The denoised signals produced by SPEAR can enable more accurate and reliable monitoring of cardiovascular health metrics from wearable devices, opening up new possibilities for personal and clinical applications of PPG technology.