Wireless Pressure Ulcer Sensor: Validation in an Animal Model.

Pressure ulcers are increasingly prevalent in an aging population. The most commonly used method of pressure ulcer prevention is pressure off-loading achieved by physically turning bedbound patients or by using expensive, single application devices such as wheelchair cushions. Our aim is to approach the problem of pressure ulcer prevention in a new way: a wireless sensor worn by the patient at locations susceptible to pressure injury. The sensor will monitor local pressure over time and transmits the data wirelessly to a base station (in a hospital setting) or smartphone (for home care). When a condition that would be harmful to tissue is reached, an alert would enable immediate direct intervention to prevent development of a pressure ulcer. The goal of this study was to validate the sensor's use in a live animal model and to lay the foundation for building time-pressure curves to predict the probability of pressure injury. Sprague-Dawley rats underwent surgical implantation of bilateral steel discs deep to the latissimus dorsi muscles. After the animals recovered from the surgical procedure, pressure was applied to the overlying tissue using magnets of varying strengths (30-150 mm Hg) for between 1 and 8 hours. Our sensor was placed on the skin prior to magnet application to wirelessly collect data regarding pressure and time. Three days after pressure application, animals were killed, injuries were graded clinically, and biopsies were collected for histological analysis. Results reveal that all animals with magnet application for more than 2 hours had clinical evidence of ulceration. Similarly, histological findings of hemorrhage were associated with increased time of pressure application. However, at high pressures (120-150 mm Hg), there were ischemic changes within the muscular layer without corresponding skin ulceration. We have developed a wireless sensor that can be placed on any at-risk area of the body and has the potential to alert caregivers when patients are at risk of developing a pressure injury. Our sensor successfully transmitted pressure readings wirelessly in a live, mobile animal. Future studies will focus on safety and efficacy with human use and development of algorithms to predict the probability of pressure ulcer formation.

Improving Pulse Rate Measurements during Random Motion Using a Wearable Multichannel Reflectance Photoplethysmograph.

Photoplethysmographic (PPG) waveforms are used to acquire pulse rate (PR) measurements from pulsatile arterial blood volume. PPG waveforms are highly susceptible to motion artifacts (MA), limiting the implementation of PR measurements in mobile physiological monitoring devices. Previous studies have shown that multichannel photoplethysmograms can successfully acquire diverse signal information during simple, repetitive motion, leading to differences in motion tolerance across channels. In this paper, we investigate the performance of a custom-built multichannel forehead-mounted photoplethysmographic sensor under a variety of intense motion artifacts. We introduce an advanced multichannel template-matching algorithm that chooses the channel with the least motion artifact to calculate PR for each time instant. We show that for a wide variety of random motion, channels respond differently to motion artifacts, and the multichannel estimate outperforms single-channel estimates in terms of motion tolerance, signal quality, and PR errors. We have acquired 31 data sets consisting of PPG waveforms corrupted by random motion and show that the accuracy of PR measurements achieved was increased by up to 2.7 bpm when the multichannel-switching algorithm was compared to individual channels. The percentage of PR measurements with error ≤ 5 bpm during motion increased by 18.9% when the multichannel switching algorithm was compared to the mean PR from all channels. Moreover, our algorithm enables automatic selection of the best signal fidelity channel at each time point among the multichannel PPG data.

A Novel Time-Varying Spectral Filtering Algorithm for Reconstruction of Motion Artifact Corrupted Heart Rate Signals During Intense Physical Activities Using a Wearable Photoplethysmogram Sensor.

Accurate estimation of heart rates from photoplethysmogram (PPG) signals during intense physical activity is a very challenging problem. This is because strenuous and high intensity exercise can result in severe motion artifacts in PPG signals, making accurate heart rate (HR) estimation difficult. In this study we investigated a novel technique to accurately reconstruct motion-corrupted PPG signals and HR based on time-varying spectral analysis. The algorithm is called Spectral filter algorithm for Motion Artifacts and heart rate reconstruction (SpaMA). The idea is to calculate the power spectral density of both PPG and accelerometer signals for each time shift of a windowed data segment. By comparing time-varying spectra of PPG and accelerometer data, those frequency peaks resulting from motion artifacts can be distinguished from the PPG spectrum. The SpaMA approach was applied to three different datasets and four types of activities: (1) training datasets from the 2015 IEEE Signal Process. Cup Database recorded from 12 subjects while performing treadmill exercise from 1 km/h to 15 km/h; (2) test datasets from the 2015 IEEE Signal Process. Cup Database recorded from 11 subjects while performing forearm and upper arm exercise. (3) Chon Lab dataset including 10 min recordings from 10 subjects during treadmill exercise. The ECG signals from all three datasets provided the reference HRs which were used to determine the accuracy of our SpaMA algorithm. The performance of the SpaMA approach was calculated by computing the mean absolute error between the estimated HR from the PPG and the reference HR from the ECG. The average estimation errors using our method on the first, second and third datasets are 0.89, 1.93 and 1.38 beats/min respectively, while the overall error on all 33 subjects is 1.86 beats/min and the performance on only treadmill experiment datasets (22 subjects) is 1.11 beats/min. Moreover, it was found that dynamics of heart rate variability can be accurately captured using the algorithm where the mean Pearson's correlation coefficient between the power spectral densities of the reference and the reconstructed heart rate time series was found to be 0.98. These results show that the SpaMA method has a potential for PPG-based HR monitoring in wearable devices for fitness tracking and health monitoring during intense physical activities.

Deep cross-modal feature learning applied to predict acutely decompensated heart failure using in-home collected electrocardiography and transthoracic bioimpedance.

BACKGROUND: Deep learning has been successfully applied to ECG data to aid in the accurate and more rapid diagnosis of acutely decompensated heart failure (ADHF). Previous applications focused primarily on classifying known ECG patterns in well-controlled clinical settings. However, this approach does not fully capitalize on the potential of deep learning, which directly learns important features without relying on a priori knowledge. In addition, deep learning applications to ECG data obtained from wearable devices have not been well studied, especially in the field of ADHF prediction. METHODS: We used ECG and transthoracic bioimpedance data from the SENTINEL-HF study, which enrolled patients (≥21 years) who were hospitalized with a primary diagnosis of heart failure or with ADHF symptoms. To build an ECG-based prediction model of ADHF, we developed a deep cross-modal feature learning pipeline, termed ECGX-Net, that utilizes raw ECG time series and transthoracic bioimpedance data from wearable devices. To extract rich features from ECG time series data, we first adopted a transfer learning approach in which ECG time series were transformed into 2D images, followed by feature extraction using ImageNet-pretrained DenseNet121/VGG19 models. After data filtering, we applied cross-modal feature learning in which a regressor was trained with ECG and transthoracic bioimpedance. Then, we concatenated the DenseNet121/VGG19 features with the regression features and used them to train a support vector machine (SVM) without bioimpedance information. RESULTS: The high-precision classifier using ECGX-Net predicted ADHF with a precision of 94 %, a recall of 79 %, and an F1-score of 0.85. The high-recall classifier with only DenseNet121 had a precision of 80 %, a recall of 98 %, and an F1-score of 0.88. We found that ECGX-Net was effective for high-precision classification, while DenseNet121 was effective for high-recall classification. CONCLUSION: We show the potential for predicting ADHF from single-channel ECG recordings obtained from outpatients, enabling timely warning signs of heart failure. Our cross-modal feature learning pipeline is expected to improve ECG-based heart failure prediction by handling the unique requirements of medical scenarios and resource limitations.

Correlation of bioimpedance changes after compressive loading of murine tissues in vivo.

OBJECTIVE: Rises in the incidence of pressure ulcers are increasingly prevalent in an aging population. Pressure ulcers are painful, are associated with increased morbidity and mortality, increase the risk for secondary infections and inpatient stay, and adds $26.8 billion annually to the healthcare costs of the USA. Evidence suggests that a change in the bioimpedance of living tissue in response to continuous local contact pressure can be a useful indicator for the onset of pressure injuries. APPROACH: Thirty-five Sprague Dawley rats were subjected to various skin pressures for differing periods of time via a surgically inserted steel disk and an externally applied magnet. Contact pressure and bioimpedance were measured and correlated with tissue loading intensity and compared to clinical ulcer grading. MAIN RESULTS: Moderate relationships between bioimpedance changes and tissue loading intensity were found. Stronger correlations were found by utilizing a combination of bioimpedance and phase angle. Thresholds were applied to the bioimpedance parameters and the usefulness of bioimpedance in classifying different ulcer stages is demonstrated. SIGNIFICANCE: These results indicate that bioimpedance may be useful as an early indicator of pressure ulcer formation and has practical significance in the development of early pressure injury detection devices.

Evaluation of key design parameters for mitigating motion artefact in the mobile reflectance PPG signal to improve estimation of arterial oxygenation.

OBJECTIVE: Pulse oximetry, a widely accepted method for non-invasive estimation of arterial oxygen saturation (SpO2) and pulse rate (PR), is increasingly being adapted for mobile applications. Previous work in mitigating motion artefact, which corrupts the photoplethysmogram (PPG) used in pulse oximetry, has focused on reducing noise using signal processing algorithms or through sensor design that controlled only one variable at a time. In this work, we have investigated the effect of several variables such as sensor weight, relative motion, placement, and contact force against the skin that can impact motion artefact independently or by interacting with each other. APPROACH: We have identified a unique combination of these variables that is most optimal in reducing motion artefacts using a full factorial design of experiments methodology and evaluated the effect of these factors on PPG readings with and without motion. MAIN RESULTS: Data collected on 10 diverse subjects showed that placement (p = 0.03), contact force (p = 0.004), and sensor-to-skin adhesion or relative motion when combined with force (p < 0.001) had the most significant effect on reducing the motion artefact signal. Sensor weight (p = 0.822) by itself had no significant effect, however when combined with sensor adhesion (p < 0.001) had a significant impact. SIGNIFICANCE: This lays the foundation for future development of more robust sensors that can significantly reduce the effect of motion artefacts in reflectance-based pulse oximetry and could have great clinical value due to significant reduction of SpO2 errors and false alarms associated with motion artefact, making wearable pulse oximetry more reliable in mobile applications.

A New Vision for Preventing Pressure Ulcers: Wearable Wireless Devices Could Help Solve a Common-and Serious-Problem.

With an aging population, the incidence and prevalence of wound problems is on the rise. Bedsores (also known as pressure ulcers or decubitus ulcers) are painful, take months to heal, and, for many patients, never do, leading to other health problems. The condition has become so acute that treating bedsores is now a significant burden on the healthcare system. An estimated 2.5 million pressure ulcers are treated in U.S. hospitals each year, adding US$11 billion annually to health care costs.

Using support vector machines on photoplethysmographic signals to discriminate between hypovolemia and euvolemia.

Identifying trauma patients at risk of imminent hemorrhagic shock is a challenging task in intraoperative and battlefield settings given the variability of traditional vital signs, such as heart rate and blood pressure, and their inability to detect blood loss at an early stage. To this end, we acquired N = 58 photoplethysmographic (PPG) recordings from both trauma patients with suspected hemorrhage admitted to the hospital, and healthy volunteers subjected to blood withdrawal of 0.9 L. We propose four features to characterize each recording: goodness of fit (r2), the slope of the trend line, percentage change, and the absolute change between amplitude estimates in the heart rate frequency range at the first and last time points. Also, we propose a machine learning algorithm to distinguish between blood loss and no blood loss. The optimal overall accuracy of discriminating between hypovolemia and euvolemia was 88.38%, while sensitivity and specificity were 88.86% and 87.90%, respectively. In addition, the proposed features and algorithm performed well even when moderate blood volume was withdrawn. The results suggest that the proposed features and algorithm are suitable for the automatic discrimination between hypovolemia and euvolemia, and can be beneficial and applicable in both intraoperative/emergency and combat casualty care.

A Robust Motion Artifact Detection Algorithm for Accurate Detection of Heart Rates From Photoplethysmographic Signals Using Time-Frequency Spectral Features.

Motion and noise artifacts (MNAs) impose limits on the usability of the photoplethysmogram (PPG), particularly in the context of ambulatory monitoring. MNAs can distort PPG, causing erroneous estimation of physiological parameters such as heart rate (HR) and arterial oxygen saturation (SpO2). In this study, we present a novel approach, "TifMA," based on using the time-frequency spectrum of PPG to first detect the MNA-corrupted data and next discard the nonusable part of the corrupted data. The term "nonusable" refers to segments of PPG data from which the HR signal cannot be recovered accurately. Two sequential classification procedures were included in the TifMA algorithm. The first classifier distinguishes between MNA-corrupted and MNA-free PPG data. Once a segment of data is deemed MNA-corrupted, the next classifier determines whether the HR can be recovered from the corrupted segment or not. A support vector machine (SVM) classifier was used to build a decision boundary for the first classification task using data segments from a training dataset. Features from time-frequency spectra of PPG were extracted to build the detection model. Five datasets were considered for evaluating TifMA performance: (1) and (2) were laboratory-controlled PPG recordings from forehead and finger pulse oximeter sensors with subjects making random movements, (3) and (4) were actual patient PPG recordings from UMass Memorial Medical Center with random free movements and (5) was a laboratory-controlled PPG recording dataset measured at the forehead while the subjects ran on a treadmill. The first dataset was used to analyze the noise sensitivity of the algorithm. Datasets 2-4 were used to evaluate the MNA detection phase of the algorithm. The results from the first phase of the algorithm (MNA detection) were compared to results from three existing MNA detection algorithms: the Hjorth, kurtosis-Shannon entropy, and time-domain variability-SVM approaches. This last is an approach recently developed in our laboratory. The proposed TifMA algorithm consistently provided higher detection rates than the other three methods, with accuracies greater than 95% for all data. Moreover, our algorithm was able to pinpoint the start and end times of the MNA with an error of less than 1 s in duration, whereas the next-best algorithm had a detection error of more than 2.2 s. The final, most challenging, dataset was collected to verify the performance of the algorithm in discriminating between corrupted data that were usable for accurate HR estimations and data that were nonusable. It was found that on average 48% of the data segments were found to have MNA, and of these, 38% could be used to provide reliable HR estimation.

Photoplethysmograph signal reconstruction based on a novel hybrid motion artifact detection-reduction approach. Part I: Motion and noise artifact detection.

Motion and noise artifacts (MNA) are a serious obstacle in utilizing photoplethysmogram (PPG) signals for real-time monitoring of vital signs. We present a MNA detection method which can provide a clean vs. corrupted decision on each successive PPG segment. For motion artifact detection, we compute four time-domain parameters: (1) standard deviation of peak-to-peak intervals (2) standard deviation of peak-to-peak amplitudes (3) standard deviation of systolic and diastolic interval ratios, and (4) mean standard deviation of pulse shape. We have adopted a support vector machine (SVM) which takes these parameters from clean and corrupted PPG signals and builds a decision boundary to classify them. We apply several distinct features of the PPG data to enhance classification performance. The algorithm we developed was verified on PPG data segments recorded by simulation, laboratory-controlled and walking/stair-climbing experiments, respectively, and we compared several well-established MNA detection methods to our proposed algorithm. All compared detection algorithms were evaluated in terms of motion artifact detection accuracy, heart rate (HR) error, and oxygen saturation (SpO2) error. For laboratory controlled finger, forehead recorded PPG data and daily-activity movement data, our proposed algorithm gives 94.4, 93.4, and 93.7% accuracies, respectively. Significant reductions in HR and SpO2 errors (2.3 bpm and 2.7%) were noted when the artifacts that were identified by SVM-MNA were removed from the original signal than without (17.3 bpm and 5.4%). The accuracy and error values of our proposed method were significantly higher and lower, respectively, than all other detection methods. Another advantage of our method is its ability to provide highly accurate onset and offset detection times of MNAs. This capability is important for an automated approach to signal reconstruction of only those data points that need to be reconstructed, which is the subject of the companion paper to this article. Finally, our MNA detection algorithm is real-time realizable as the computational speed on the 7-s PPG data segment was found to be only 7 ms with a Matlab code.

Photoplethysmograph signal reconstruction based on a novel motion artifact detection-reduction approach. Part II: Motion and noise artifact removal.

We introduce a new method to reconstruct motion and noise artifact (MNA) contaminated photoplethysmogram (PPG) data. A method to detect MNA corrupted data is provided in a companion paper. Our reconstruction algorithm is based on an iterative motion artifact removal (IMAR) approach, which utilizes the singular spectral analysis algorithm to remove MNA artifacts so that the most accurate estimates of uncorrupted heart rates (HRs) and arterial oxygen saturation (SpO2) values recorded by a pulse oximeter can be derived. Using both computer simulations and three different experimental data sets, we show that the proposed IMAR approach can reliably reconstruct MNA corrupted data segments, as the estimated HR and SpO2 values do not significantly deviate from the uncorrupted reference measurements. Comparison of the accuracy of reconstruction of the MNA corrupted data segments between our IMAR approach and the time-domain independent component analysis (TD-ICA) is made for all data sets as the latter method has been shown to provide good performance. For simulated data, there were no significant differences in the reconstructed HR and SpO2 values starting from 10 dB down to -15 dB for both white and colored noise contaminated PPG data using IMAR; for TD-ICA, significant differences were observed starting at 10 dB. Two experimental PPG data sets were created with contrived MNA by having subjects perform random forehead and rapid side-to-side finger movements show that; the performance of the IMAR approach on these data sets was quite accurate as non-significant differences in the reconstructed HR and SpO2 were found compared to non-contaminated reference values, in most subjects. In comparison, the accuracy of the TD-ICA was poor as there were significant differences in reconstructed HR and SpO2 values in most subjects. For non-contrived MNA corrupted PPG data, which were collected with subjects performing walking and stair climbing tasks, the IMAR significantly outperformed TD-ICA as the former method provided HR and SpO2 values that were non-significantly different than MNA free reference values.

Physiological parameter monitoring from optical recordings with a mobile phone.

We show that a mobile phone can serve as an accurate monitor for several physiological variables, based on its ability to record and analyze the varying color signals of a fingertip placed in contact with its optical sensor. We confirm the accuracy of measurements of breathing rate, cardiac R-R intervals, and blood oxygen saturation, by comparisons to standard methods for making such measurements (respiration belts, ECGs, and pulse-oximeters, respectively). Measurement of respiratory rate uses a previously reported algorithm developed for use with a pulse-oximeter, based on amplitude and frequency modulation sequences within the light signal. We note that this technology can also be used with recently developed algorithms for detection of atrial fibrillation or blood loss.

Statistical approach for the detection of motion/noise artifacts in Photoplethysmogram.

Motion and noise artifacts (MNA) have been a serious obstacle in realizing the potential of Photoplethysmogram (PPG) signals for real-time monitoring of vital signs. We present a statistical approach based on the computation of kurtosis and Shannon Entropy (SE) for the accurate detection of MNA in PPG data. The MNA detection algorithm was verified on multi-site PPG data collected from both laboratory and clinical settings. The accuracy of the fusion of kurtosis and SE metrics for the artifact detection was 99.0%, 94.8% and 93.3% in simultaneously recorded ear, finger and forehead PPGs obtained in a clinical setting, respectively. For laboratory PPG data recorded from a finger with contrived artifacts, the accuracy was 88.8%. It was identified that the measurements from the forehead PPG sensor contained the most artifacts followed by finger and ear. The proposed MNA algorithm can be implemented in real-time as the computation time was 0.14 seconds using Matlab®.

A novel approach using time-frequency analysis of pulse-oximeter data to detect progressive hypovolemia in spontaneously breathing healthy subjects.

Accurate and early detection of blood volume loss would greatly improve intraoperative and trauma care. This study has attempted to determine early diagnostic and quantitative markers for blood volume loss by analyzing photoplethysmogram (PPG) data from ear, finger and forehead sites with our high-resolution time-frequency spectral (TFS) technique in spontaneously breathing healthy subjects (n = 11) subjected to lower body negative pressure (LBNP). The instantaneous amplitude modulations present in heart rate (AM HR) and breathing rate (AMBR) band frequencies of PPG signals were calculated from the high-resolution TFS. Results suggested that the changes (P < 0.05) in AMBR and especially in AMHR values can be used to detect the blood volume loss at an early stage of 20% LBNP tolerance when compared to the baseline values. The mean percent decrease in AMHR values at 100% LBNP tolerance was 78.3%, 72.5%, and 33.9% for ear, finger, and forehead PPG signals, respectively. The mean percent increase in AMBR values at 100% LBNP tolerance was 99.4% and 19.6% for ear and finger sites, respectively; AMBR values were not attainable for forehead PPG signal. Even without baseline AMHR values, our results suggest that hypovolemia detection is possible with specificity and sensitivity greater than 90% for the ear and forehead locations when LBNP tolerance is 100%. Therefore, the TFS analysis of noninvasive PPG waveforms is promising for early diagnosis and quantification of hypovolemia at levels not identified by vital signs in spontaneously breathing subjects.

A comparative evaluation of adaptive noise cancellation algorithms for minimizing motion artifacts in a forehead-mounted wearable pulse oximeter.

Wearable physiological monitoring using a pulse oximeter would enable field medics to monitor multiple injuries simultaneously, thereby prioritizing medical intervention when resources are limited. However, a primary factor limiting the accuracy of pulse oximetry is poor signal-to-noise ratio since photoplethysmographic (PPG) signals, from which arterial oxygen saturation (SpO2) and heart rate (HR) measurements are derived, are compromised by movement artifacts. This study was undertaken to quantify SpO2 and HR errors induced by certain motion artifacts utilizing accelerometry-based adaptive noise cancellation (ANC). Since the fingers are generally more vulnerable to motion artifacts, measurements were performed using a custom forehead-mounted wearable pulse oximeter developed for real-time remote physiological monitoring and triage applications. This study revealed that processing motion-corrupted PPG signals by least mean squares (LMS) and recursive least squares (RLS) algorithms can be effective to reduce SpO2 and HR errors during jogging, but the degree of improvement depends on filter order. Although both algorithms produced similar improvements, implementing the adaptive LMS algorithm is advantageous since it requires significantly less operations.

Reflectance forehead pulse oximetry: effects of contact pressure during walking.

Steady progress has been made towards the development of a reliable wearable pulse oximeter to aid first responders in remote monitoring and triage operations. This study was undertaken to assess how varying contact pressures affects the photoplethysmographic (PPG) signal, and arterial oxygen saturation (SpO2) and heart rate (HR) measurement errors during motion artifact inducing activity. The study revealed that contact pressures ranging from 8-12 kPa resulted in the largest PPG amplitude for a reflectance sensor attached to the forehead region above the eye, although the signal-to-noise ratio (SNR) did not improve significantly. However, SpO2 and HR errors increased when insufficient contact pressure was applied. This information may be helpful in the design of a more robust pulse oximeter sensor for use in remote monitoring applications.

Multi-wavelength reflectance pulse oximetry.

The performance of current reflectance pulse oximeters is hindered by poor signal-to-noise ratio. To overcome this problem a new reflectance oximeter has been developed with a sensor which consists of three LEDs and two continuous photodetector rings placed equidistant from the center of the LEDs. In addition, ultra low noise electronics and adaptive algorithm assure improved performance. A validation study was performed on 10 healthy volunteers. Sensors were placed on several sites and measurements were compared to reference arterial blood samples. During the study progressive hypoxemia was induced by lowering the inspired oxygen concentration (FiO2) to 10%, followed by a recovery phase. Twelve blood samples were taken during each cycle, yielding a total of 120 measured data points. Data from randomly selected 5 subjects was used for calibration and subsequently tested on the other 5 subjects. Results proved to be well within clinically acceptable boundaries for all 3 sampling sites with high correlation (R2 > 0.9) and SD around 2%. In conclusion, a new 3 wavelength reflectance pulse oximeter with unique sensor geometry and improved algorithms provides enhanced performance and is less susceptible to poor signal to noise conditions when compared to existing reflectance oximetry systems.