Multiparameter Intelligent Monitoring In Intensive Care Research Articles

Comparative study on the predictive value of TG/HDL-C, TyG and TyG-BMI indices for 5-year mortality in critically ill patients with chronic heart failure: a retrospective study

BackgroundThe triglyceride glucose (TyG) index, TyG-body mass index (TyG-BMI), and triglyceride-density lipoprotein cholesterol ratio (TG/HDL-C) are substitute indicators for insulin resistance (IR). This study aimed to compare the predictive value of these indicators for 5-year mortality in critically ill patients with chronic heart failure (CHF).MethodsCritically ill patients with CHF were identified from the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) III and IV databases. The primary outcome was 5-year mortality. The relationship between the three indices and mortality risk was determined using multivariate Cox proportional hazards models, Kaplan–Meier (K‒M) analysis and restricted cubic splines analysis. A receiver operating characteristic (ROC) curve was generated to compare the ability of the three indices to predict mortality. Finally, whether the IR indices would further increase the predictive ability of the basic model including baseline variables with a significance level between survivors and non-survivors was evaluated by ROC curve.ResultsAltogether, 1329 patients with CHF were identified from the databases. Cox proportional hazards models indicated that the TyG index was independently associated with an elevated risk of 5-year mortality (hazard ratio [HR], 1.56; 95% confidence interval [CI] 1.29–1.9), while the TyG-BMI index and TG/HDL-C level were significantly associated with 5-year mortality, with an HR (95% CI) of 1.002 (1.000–1.003) and 1.01 (1.00–1.03), respectively. The K–M analysis revealed that the cumulative incidence of all-cause 5-year death increased with increasing quartiles of the TyG index, TyG-BMI index, or TG/HDL-C ratio. According to the ROC curve, the TyG index outperformed the TyG-BMI and TG/HDL-C ratio at predicting all-cause 5-year mortality (0.608 [0.571–0.645] vs. 0.558 [0.522–0.594] vs. 0.561 [0.524–0.598]). The effect of the TyG index on all-cause mortality was consistent across subgroups, with no significant interaction with randomized factors. Furthermore, adding the TyG index to the basic model for 5-year mortality improved its predictive ability (area under the curve, 0.762 for the basic model vs. 0.769 for the basic model + TyG index); however, the difference was not statistically significant.ConclusionAs continuous variables, all three indices were significantly associated with 5-year mortality risk in critically ill patients with CHF. Although these IR indices did not improve the predictive power of the basic model in patients with CHF, the TyG index appears to be the most promising index (vs. TyG-BMI and TG/HDL-C ratio) for prevention and risk stratification in critically ill patients with CHF.

Predictive value of admission red cell distribution width-to-platelet ratio for 30-day death in patients with spontaneous intracerebral hemorrhage: an analysis of the MIMIC database.

Prognostic assessment plays an important role in the effective management of patients with spontaneous intracerebral hemorrhage (ICH). The study aimed to investigate whether elevated red cell distribution width-to-platelet ratio (RPR) at admission was related to 30-day death in patients with spontaneous intracerebral hemorrhage (ICH). This retrospective cohort study included 2,823 adult patients with ICH from the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) III and IV databases between 2001 and 2019. The Cox proportional hazard model was utilized to evaluate the relationship between RPR levels and 30-day death risk. The area under receiver-operating characteristic curve (AUC) was used to assess the predictive ability of RPR for 30-day death in patients with ICH. At the end of the 30-day follow-up, 799 (28.30%) patients died, and the median RPR level was 0.066 (0.053, 0.087). After adjusting for confounders, the tertile 3 of RPR levels [hazard ratio (HR) = 1.37, 95% confidence interval (CI): 1.15-1.64] were associated with a higher risk of 30-day death in patients with ICH compared with tertile 1. In the stratified analyses, elevated RPR levels were found to be associated with an increased risk of 30-day death in patients aged <65 years (HR = 1.77, 95%CI: 1.29-2.43), aged ≥65 years (HR = 1.30, 95%CI: 1.05-1.61), with Glasgow Coma Score (GCS) <14 (HR = 1.65, 95%CI: 1.27-2.14), with Charlson comorbidity index (CCI) ≥4 (HR = 1.45, 95%CI: 1.17-1.80), with (HR = 1.66, 95%CI: 1.13-2.43) or without sepsis (HR = 1.32, 95%CI: 1.08-1.61), and female patients (HR = 1.75, 95%CI: 1.35-2.26) but not in male patients (P = 0.139) and patients with GCS ≥14 (P = 0.058) or CCI <4 (P = 0.188). The AUC for RPR to predict 30-day death in patients with ICH was 0.795 (95%CI: 0.763-0.828) in the testing set, indicating a good predictive ability. Elevated RPR levels were correlated with an increased risk of 30-day death in patients with ICH, and RPP levels showed good predictive ability for 30-day death.

Fluid and Solute Intakes Show Minimal Association With Serum Sodium Levels in a Mixed ICU Population.

Hyponatremia is common among hospital inpatients. It is generally due to excess free body water resulting from increased water intake and decreased water elimination due to underlying pathology and hormonal influence. However, supporting evidence is lacking for treating mild hyponatremia with fluid restriction. Our study examines the association between hyponatremia and fluid intake in acutely ill inpatients. We hypothesize that fluid intake is not closely associated with serum sodium (SNa). We conducted a retrospective study of hyponatremia using the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) III dataset, a public ICU registry. We analyzed fluid, sodium, and potassium intake with a mixed model linear regression with SNa as the outcome for hyponatremic and non-hyponatremic patients and cumulative total input from one to seven days. In addition, we compared a group of patients receiving less than one liter of fluid per day to a group receiving more than one liter. The association of SNa with fluid intake was negative and statistically significant for most cumulative days of intake from one to seven for the total population and those with sporadic hyponatremia. For those with uniform hyponatremia, the negative association was significant for three and four days of cumulative input. The change in SNa was almost always less than 1 mmol/L of additional fluid intake across all groups. SNa for hyponatremic patients who received less than one liter of fluid per day were within one mmol/L of those who received more (p<0.001 for one, two, and seven cumulative intake days). SNa is associated with a change of less than 1 mmol/L across a wide range of fluid and sodium intake in adult ICU patients. Patients who received less than one liter per day had SNa almost identical to those who received more. This suggests that SNa is not tightly coupled with fluid intake in the acutely ill population and that hormonal control of water elimination is the predominant mechanism. This might explain why the correction of hyponatremia by fluid restriction is often difficult.

Detecting Atrial Fibrillation in Real Time Based on PPG via Two CNNs for Quality Assessment and Detection

Real-time detection of atrial fibrillation (AFib) is made possible by the quality assessment via a 1-D convolutional neural network (1D-CNN) in a processor of a photoplethysmography (PPG) sensor patch and a 2-D convolutional neural network (2D-CNN) for AFib detection in cloud. The 1D-CNN is able to filter out the unqualified PPG that is contaminated by motion artifacts and/or ambient light interference. The remaining qualified PPG is then inputted to another built 2D-CNN for detecting AFib. This 1D-CNN consists of four layers of convolutions and max pooling, one long short-term memory (LSTM), and an output dense layer. The 2D-CNN is pretrained based on the electrocardiography (ECG) data from multiparameter intelligent monitoring in intensive care (MIMIC) III database, for which the RR-intervals (RRIs) of ECG data are first extracted in Poincaré images and then regarded as input features to the model for training. This 2D-CNN has also four layers of convolutions and max pooling and four output dense layers. The pretrained model is next fine-tuned based on peak-to-peak intervals (PPIs) of PPG measured from wearable devices as input features for detecting AFib effectively. The quality-assessment 1D-CNN model is implemented in the wearable device to transmit only qualified data to the 2D-CNN model in cloud for AFib detection, achieving power efficiency. Both models are trained by the Adam optimizer. To validate the models, the PPIs of PPG were collected to evaluate the performance of the established models in real time. Experimental results show that the fine-tuned 2D-CNN for AFib detection achieves the accuracy, sensitivity, and specificity were 98.08%, 96.82%, and 98.86%, respectively, the most favorable as opposed to other reported works based on PPG. The models are able to not only assist clinicians in AFib detection but also provide a mechanism to detect AFib via wearable devices in real time.

Higher Order Derivative-Based Integrated Model for Cuff-Less Blood Pressure Estimation and Stratification Using PPG Signals

Precise blood pressure (BP) estimation is vital for diagnosing arterial hypertension and other cardiovascular ailments. The photoplethysmogram (PPG)-based cuff-less BP measurement is an alternative to the traditional cuff-based systems. The morphological, temporal, and frequency-domain-based features have been used for BP estimation in the PPG-based BP measurement systems. However, dealing with varying signal morphology and feature dependency on the fiducial points in PPG contours remains challenging and limits the performance of the existing BP estimation algorithms, especially in wearable devices. This work presents a novel approach that considers the nonlinear features of PPG signals evaluated using higher order derivatives. In particular, the PPG signal’s third and fourth derivative contours are used to extract features, such as fractal dimension, bubble entropy (BE), Lyapunov exponent, and moving slope. Machine learning algorithms such as random forest (RF), extreme gradient boosting (XGBoost), and support vector regression (SVR) models are used for BP estimation using the nonlinear features of PPG signals. The estimated BP values are further categorized based on five broad classes based on the BP stratification criteria such as hypotension, normal, prehypertension, stage-I, and stage-II hypertension, respectively. The performance of the suggested approach is evaluated using PPG signals from three publicly available databases (multiparameter intelligent monitoring in intensive care (MIMIC)-I, II, and III). The proposed estimation approach has outperformed the recent existing algorithms and achieved a minimum value of mean absolute error (MAE) ± standard deviation (STD) in (systolic BP (SBP) and diastolic BP (DBP) values) as <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${0}.{74}\pm {2}.{42}$ </tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${0}.{35} \pm {1}.{06}$ </tex-math></inline-formula> , respectively, for the MIMIC-I database. The suggested approach has also achieved grade-A on the British Hypertension Society (BHS) standard.

Comparative Efficacy of Fentanyl and Morphine in Patients with or At Risk for Acute Respiratory Distress Syndrome: A Propensity Score-Matched Cohort Study

IntroductionOpioids are potent painkillers but can have severe adverse effects in the intensive care unit (ICU). The aim of this study was to compare the outcomes of fentanyl and morphine use among patients at risk for and with acute respiratory distress syndrome (ARDS).MethodsWe developed a dataset of real-world data to enable the comparison of the effectiveness and safety of opioids and the associated outcomes from the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC)-III database and the eICU Collaborative Research Database. Patients who were admitted to the ICU with a diagnosis of or at risk for ARDS and received mechanical ventilation for at least 12 h were included. Patients were enrolled sequentially into one of six groups in three cohorts: treated with fentanyl or not; treated with morphine or not; and treated with fentanyl or morphine. Propensity score matching and multivariable analyses were performed.ResultsFentanyl was associated with higher in-hospital mortality in the propensity score-matched model but not in the linear regression model. The use of morphine was associated with a higher in-hospital mortality in both models. Both fentanyl and morphine were associated with longer duration of mechanical ventilation, ICU stay, and hospitalization and a decreased likelihood of being discharged home in both models. Notably, compared with morphine, fentanyl was associated with a lower mortality and an increased likelihood of being discharged home.ConclusionsBoth fentanyl and morphine were independent risk factors for worse outcomes in patients with or at risk for ARDS. Compared with morphine, fentanyl may be preferred in these patients.Supplementary InformationThe online version contains supplementary material available at 10.1007/s40268-021-00338-3.

An Adaptive Weight Learning-Based Multitask Deep Network for Continuous Blood Pressure Estimation Using Electrocardiogram Signals.

Estimating blood pressure via combination analysis with electrocardiogram and photoplethysmography signals has attracted growing interest in continuous monitoring patients’ health conditions. However, most wearable/portal monitoring devices generally acquire only one kind of physiological signals due to the consideration of energy cost, device weight and size, etc. In this study, a novel adaptive weight learning-based multitask deep learning framework based on single lead electrocardiogram signals is proposed for continuous blood pressure estimation. Specifically, the proposed method utilizes a 2-layer bidirectional long short-term memory network as the sharing layer, followed by three identical architectures of 2-layer fully connected networks for task-specific blood pressure estimation. To learn the importance of task-specific losses automatically, an adaptive weight learning scheme based on the trend of validation loss is proposed. Extensive experiment results on Physionet Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) II waveform database demonstrate that the proposed method using electrocardiogram signals obtains estimating performance of mmHg, mmHg, and mmHg for systolic blood pressure, diastolic blood pressure, and mean arterial pressure, respectively. It can meet the requirements of the British Hypertension Society standard and US Association of Advancement of Medical Instrumentation standard with a considerable margin. Combined with a wearable/portal electrocardiogram device, the proposed model can be deployed to a healthcare system to provide a long-term continuous blood pressure monitoring service, which would help to reduce the incidence of malignant complications to hypertension.

An Efficient Fingertip Photoplethysmographic Signal Artifact Detection Method: A Machine Learning Approach

A photoplethysmography method has recently been widely used to noninvasively measure blood volume changes during a cardiac cycle. Photoplethysmogram (PPG) signals are sensitive to artifacts that negatively impact the accuracy of many important measurements. In this paper, we propose an efficient system for detecting PPG signal artifacts acquired from a fingertip in the public healthcare database named Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) by using 11 features as the input of the random forest algorithm and classified the signals into two classes: acceptable and anomalous. A real‐time algorithm is proposed to identify artifacts by using the method. The efficient Fisher score feature selection algorithm was used to order and select 11 relevant features from 19 available features that represented the PPG signal very effectively. Six machine learning algorithms (random forest, decision tree, Gaussian naïve Bayes, linear support vector machine, artificial neural network, and probabilistic neural network) were applied with the extracted features, and their classification accuracy was measured. Among them, the random forest had the best performance using only 11 out of 19 features with an accuracy of 85.68%. Our proposed method also achieved good sensitivity and specificity value of 86.57% and 85.09%, respectively. The proposed real‐time algorithm can be an easy and convenient way for real‐time PPG signal artifact detection using smartphones and wearable devices.

Lactate indices as predictors of in-hospital mortality or 90-day survival after admission to an intensive care unit in unselected critically ill patients.

BackgroundWe performed an exclusive study to investigate the associations between a total of 23 lactate-related indices during the first 24h in an intensive care unit (ICU) and in-hospital mortality.MethodsNine static and 14 dynamic lactate indices, including changes in lactate concentrations (Δ Lac) and slope (linear regression coefficient), were calculated from individual critically ill patient data extracted from the Multiparameter Intelligent Monitoring for Intensive Care (MIMIC) III database.ResultsData from a total of 781 ICU patients were extracted, consisted of 523 survivors and 258 non-survivors. The in-hospital mortality rate for this cohort was 33.0%. A multivariate logistic regression model identified maximal lactate concentration at 24h after ICU admission (max lactate at T24) as a significant predictor of in-hospital mortality (odds ratio = 1.431, 95% confidence interval [CI] = 1.278–1.604, p<0.001) after adjusting for predefined confounders (age, gender, sepsis, Elixhauser comorbidity score, mechanical ventilation, renal replacement therapy, vasopressors, ICU severity scores). Area under curve (AUC) for max lactate at T24 was larger (AUC = 0.776, 95% CI = 0.740–0.812) than other indices (p<0.001), comparable to an APACHE III score of 0.771. When combining max lactate at T24 with APACHE III, the AUC was increased to 0.815 (95% CI:0.783–0.847). The sensitivity, specificity, and positive and negative predictive values for the cut-off value of 3.05 mmol/L were 64.3%, 77.4%, 58.5%, and 81.5%, respectively.Kaplan-Myer survival curves of the max lactate at T24 for 90-day survival after admission to ICU demonstrated a significant difference according to the cut-off value (p<0.001).ConclusionsThese data indicate that the maximal arterial lactate concentration at T24 is a robust predictor of in-hospital mortality as well as 90-day survival in unselected ICU patients with predictive ability as comparable with APACHE III score.

Intelligent applications of cloud computing in enhancing health care services

In this work, a thorough evaluation of an approach to extract protocols in a series of time data has been presented based on Biological parameters in medical situations. This methodology is wholly data-enhanced with its prototypical pattern mined during every Biological timeframe dataset. The produced rules, centered on the prototypical design have been illustrated in their connection to the wide range of prototypical data arrangements. With that regard, this paper provides an analysis of an approach to effectively measure the similarity of protocol arrangements which have been introduced to verify the uniqueness of protocol arrangements. This approach is evaluated based on Biological records from the medical category in the Multi-parameter Intelligent Monitoring for Intensive Care (MIMIC) online databases such as respiratory failure, sepsis, or angina. The findings in this analysis indicate that data mining approaches can receive a distinctive framework for every medical situation and represent the produced mining rules in an understandable and contextual format.

Identification key scheme to enhance network performance in wireless body area network

Wireless Body Area Network (WBAN) refers to a group of small intelligent electronic devices placed on the human body to monitor its vital signals. the efficiency of network performance in health care applications is one of the most important prerequisites for designing an effective surveillance system. However, none of the previous studies have succeeded in proposing an effective scheme that could cover the requirements for effective network performance or the development of the WBAN system capable of sending and receiving vital patient data efficiently. The aim of this paper is to enhance the performance of the network in WBAN. In addition, to verify the transfer of patient data from sensors on the body to receivers with the least value of possible packet loss, a solution is developed that can reduce the value of packet loss, end to end delay and increase the value of throughput, relying on confirmation and matching data between the sender and receiver devices. A new scheme known as the Identification Key Scheme (IKS) is proposed based on the newly available 402–405 MHz Medical Implant Communication Service (MICS). The analysis of the network performance was implemented in real WBAN dataset and used the Multi-Parameter Intelligent Monitoring for Intensive Care (MIMIC) database. MIMIC database included the data recorded from 90 Intensive Care Unit (ICU) patients. In this study, OMNET++ was used as the network simulation tool to design and evaluate the proposed scheme. The results showed that IKS was able to reduce number of packet loss and end to end delay with increased throughput. This is an indicator of the ability of IKS in providing efficient network performance between devices in the WBAN application.

Automated Screening of Myocardial Infarction Based on Statistical Analysis of Photoplethysmographic Data

Integration of automated cardiac analysis techniques with the personal healthcare devices can aid in the timely diagnosis of myocardial infarction (MI) to reduce the risk of mortality. The use of electrocardiogram (ECG) along with advanced signal processing tools limits the application of the available techniques for such portable devices having limited processing ability. Here we propose a novel MI screening technique based on photoplethysmographic (PPG) data, which is the most widely monitored signal in everyday health monitoring. Two simple statistical features ideally reflecting the MI-induced variability in the PPG data are identified. To combat the noise sensitivity of the PPG acquisition process, the same statistical parameters are employed for eliminating the noise corrupted data segments. Classification of MI data is then performed using a simple threshold-based binary classification approach. The technique is evaluated with PPG records collected from 62 hospitalized MI patients and 65 healthy subjects. The robustness of the algorithm is cross-validated with both MI and non-MI records collected from the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) II database. The algorithm achieves an average detection accuracy of 96.25% with sensitivity and specificity of 96.36% and 96.15%, respectively. The promising results establish the utility of PPG as an alternate MI diagnosing tool. Apart from using the less obtrusive PPG signal, the proposed technique relies on simple feature extraction and classification tools, thus making it a promising alternative to the ECG based techniques for implementation on the personal healthcare devices.

Identifying Clinical Terms in Medical Text Using Ontology-Guided Machine Learning.

BackgroundAutomatic recognition of medical concepts in unstructured text is an important component of many clinical and research applications, and its accuracy has a large impact on electronic health record analysis. The mining of medical concepts is complicated by the broad use of synonyms and nonstandard terms in medical documents.ObjectiveWe present a machine learning model for concept recognition in large unstructured text, which optimizes the use of ontological structures and can identify previously unobserved synonyms for concepts in the ontology.MethodsWe present a neural dictionary model that can be used to predict if a phrase is synonymous to a concept in a reference ontology. Our model, called the Neural Concept Recognizer (NCR), uses a convolutional neural network to encode input phrases and then rank medical concepts based on the similarity in that space. It uses the hierarchical structure provided by the biomedical ontology as an implicit prior embedding to better learn embedding of various terms. We trained our model on two biomedical ontologies—the Human Phenotype Ontology (HPO) and Systematized Nomenclature of Medicine - Clinical Terms (SNOMED-CT).ResultsWe tested our model trained on HPO by using two different data sets: 288 annotated PubMed abstracts and 39 clinical reports. We achieved 1.7%-3% higher F1-scores than those for our strongest manually engineered rule-based baselines (P=.003). We also tested our model trained on the SNOMED-CT by using 2000 Intensive Care Unit discharge summaries from MIMIC (Multiparameter Intelligent Monitoring in Intensive Care) and achieved 0.9%-1.3% higher F1-scores than those of our baseline. The results of our experiments show high accuracy of our model as well as the value of using the taxonomy structure of the ontology in concept recognition.ConclusionMost popular medical concept recognizers rely on rule-based models, which cannot generalize well to unseen synonyms. In addition, most machine learning methods typically require large corpora of annotated text that cover all classes of concepts, which can be extremely difficult to obtain for biomedical ontologies. Without relying on large-scale labeled training data or requiring any custom training, our model can be efficiently generalized to new synonyms and performs as well or better than state-of-the-art methods custom built for specific ontologies.

Photoplethysmography and Deep Learning: Enhancing Hypertension Risk Stratification.

Blood pressure is a basic physiological parameter in the cardiovascular circulatory system. Long-term abnormal blood pressure will lead to various cardiovascular diseases, making the early detection and assessment of hypertension profoundly significant for the prevention and treatment of cardiovascular diseases. In this paper, we investigate whether or not deep learning can provide better results for hypertension risk stratification when compared to the classical signal processing and feature extraction methods. We tested a deep learning method for the classification and evaluation of hypertension using photoplethysmography (PPG) signals based on the continuous wavelet transform (using Morse) and pretrained convolutional neural network (using GoogLeNet). We collected 121 data recordings from the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) Database, each containing arterial blood pressure (ABP) and photoplethysmography (PPG) signals. The ABP signals were utilized to extract blood pressure category labels, and the PPG signals were used to train and test the model. According to the seventh report of the Joint National Committee, blood pressure levels are categorized as normotension (NT), prehypertension (PHT), and hypertension (HT). For the early diagnosis and assessment of HT, the timely detection of PHT and the accurate diagnosis of HT are significant. Therefore, three HT classification trials were set: NT vs. PHT, NT vs. HT, and (NT + PHT) vs. HT. The F-scores of these three classification trials were 80.52%, 92.55%, and 82.95%, respectively. The tested deep method achieved higher accuracy for hypertension risk stratification when compared to the classical signal processing and feature extraction method. Additionally, the method achieved comparable results to another approach that requires electrocardiogram and PPG signals.

PQR signal quality indexes: A method for real-time photoplethysmogram signal quality estimation based on noise interferences

Abstract The photoplethysmograph (PPG) acquired from pulse oximeters has been extensively used to estimate the heart rate and blood oxygen saturation. However, how to improve the accuracy of these parameters is always a challenging work as PPG is susceptible to be contaminated by noises especially motion artifacts (MAs). This study presents an algorithm called PQR to calculate three indexes (P/Q/R) for estimation the signal quality based on noise interferences and rSQI index for comparison of any two signals. P/Q/R indicates the influencing degree of high-frequency noise, baseline wandering and motion artifacts on PPG respectively. And a relative signal quality estimation index rSQI is designed to compare the quality of any two signals. When the value of rSQI is greater than zero, it illustrates that the quality of the former signal is better than the latter, and vice versa. Lower P/Q/R scores could be obtained from the algorithm if PPG is contaminated by strong artifacts or with an irregular signal morphology. The algorithm put a complete successive signal into consideration instead of signal segmentation beat by beat. Experiment results using MIMIC (Multi-parameter Intelligent Monitoring for Intensive care) database indicates the availability of PQR algorithm. And the real-time PPG signal quality estimation system could be realized to help estimating heart rate online by applying PQR algorithm to the dataset of IEEE Signal Processing Society. Moreover, the experiments using our lab hardware platform show the corresponding preliminary results of application on the comparison of different signal processing methods. The novel PQR algorithm generates significantly lower rSQI values with the comparison of reference signal when PPG signals are contaminated by strong noise. The chosen threshold could be established once an application has been defined derived from the collected data. The PQR algorithm maybe could help to guide the choice of PPG signal in real-time for the physical information extraction such as heart rate.

A novel Blood Pressure estimation method combing Pulse Wave Transit Time model and neural network model.

Blood Pressure (BP) measurement can assist doctors to assess patients' cardiovascular status and diagnose heart diseases. Pulse Wave Transit Time (PWTT) model is one frequently used BP estimation method to monitor BP continuously in clinics. However, individual variations may influence the measurement accuracy of PWTT model. Focusing on above promble, this paper proposes a novel BP estimation method combining a classical PWTT model and a neural network model. The novel method is composed of five steps: signal pre-processing, feature extraction, initial PWTT model selection, model correction by neural network model, and final PWTT model identification. A validation experiment based on 10 patients from Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) database showed that the BP estimation results by our method had a minimum mean of error readout value 5 mmHg with a standard deviation of error readout value ±8mmHg. As a result, both the diastolic blood pressure and systolic blood pressure estimation by our method can meet clinical requirements.

Prediction of Sepsis in the Intensive Care Unit With Minimal Electronic Health Record Data: A Machine Learning Approach

BackgroundSepsis is one of the leading causes of mortality in hospitalized patients. Despite this fact, a reliable means of predicting sepsis onset remains elusive. Early and accurate sepsis onset predictions could allow more aggressive and targeted therapy while maintaining antimicrobial stewardship. Existing detection methods suffer from low performance and often require time-consuming laboratory test results.ObjectiveTo study and validate a sepsis prediction method, InSight, for the new Sepsis-3 definitions in retrospective data, make predictions using a minimal set of variables from within the electronic health record data, compare the performance of this approach with existing scoring systems, and investigate the effects of data sparsity on InSight performance.MethodsWe apply InSight, a machine learning classification system that uses multivariable combinations of easily obtained patient data (vitals, peripheral capillary oxygen saturation, Glasgow Coma Score, and age), to predict sepsis using the retrospective Multiparameter Intelligent Monitoring in Intensive Care (MIMIC)-III dataset, restricted to intensive care unit (ICU) patients aged 15 years or more. Following the Sepsis-3 definitions of the sepsis syndrome, we compare the classification performance of InSight versus quick sequential organ failure assessment (qSOFA), modified early warning score (MEWS), systemic inflammatory response syndrome (SIRS), simplified acute physiology score (SAPS) II, and sequential organ failure assessment (SOFA) to determine whether or not patients will become septic at a fixed period of time before onset. We also test the robustness of the InSight system to random deletion of individual input observations.ResultsIn a test dataset with 11.3% sepsis prevalence, InSight produced superior classification performance compared with the alternative scores as measured by area under the receiver operating characteristic curves (AUROC) and area under precision-recall curves (APR). In detection of sepsis onset, InSight attains AUROC = 0.880 (SD 0.006) at onset time and APR = 0.595 (SD 0.016), both of which are superior to the performance attained by SIRS (AUROC: 0.609; APR: 0.160), qSOFA (AUROC: 0.772; APR: 0.277), and MEWS (AUROC: 0.803; APR: 0.327) computed concurrently, as well as SAPS II (AUROC: 0.700; APR: 0.225) and SOFA (AUROC: 0.725; APR: 0.284) computed at admission (P<.001 for all comparisons). Similar results are observed for 1-4 hours preceding sepsis onset. In experiments where approximately 60% of input data are deleted at random, InSight attains an AUROC of 0.781 (SD 0.013) and APR of 0.401 (SD 0.015) at sepsis onset time. Even with 60% of data missing, InSight remains superior to the corresponding SIRS scores (AUROC and APR, P<.001), qSOFA scores (P=.0095; P<.001) and superior to SOFA and SAPS II computed at admission (AUROC and APR, P<.001), where all of these comparison scores (except InSight) are computed without data deletion.ConclusionsDespite using little more than vitals, InSight is an effective tool for predicting sepsis onset and performs well even with randomly missing data.

A NOVEL CONTINUOUS AND NONINVASIVE MEASUREMENT FOR BLOOD PRESSURE BASED ON PHOTOPLETHYSMOGRAPHY

Implementation of continuous noninvasive blood pressure (NIBP) measurement is important for prevention and control of the cardiovascular disease. This paper study a novel continuous noninvasive blood pressure (BP) fitting method based on the photoplethysmography (PPG), electrocardiogram (ECG) and blood volume characteristics. The estimation model of systolic blood pressure (SBP) was established from ECG and PPG in Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) II openness medical database. According to analysis of Beer–Lambert law applied in arterial blood characteristics and PPG waveform, the estimation model of differential pressure (DP) was constructed. Taking the SBP and diastolic blood pressure (DBP) from MIMIC II database as the gold standard value, the validation results show that the mean difference and standard deviation of the fitting results are within the standard range of ([Formula: see text])[Formula: see text]mm/Hg. The method proposed in this paper can effectively achieve continuous BP measurement.

Discharge recommendation based on a novel technique of homeostatic analysis.

We propose a computational framework for integrating diverse patient measurements into an aggregate health score and applying it to patient stability prediction. We mapped retrospective patient data from the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC) II clinical database into a discrete multidimensional space, which was searched for measurement combinations and trends relevant to patient outcomes of interest. Patient trajectories through this space were then used to make outcome predictions. As a case study, we built AutoTriage, a patient stability prediction tool to be used for discharge recommendation. AutoTriage correctly identified 3 times as many stabilizing patients as existing tools and achieved an accuracy of 92.9% (95% CI: 91.6-93.9%), while maintaining 94.5% specificity. Analysis of AutoTriage parameters revealed that interdependencies between risk factors comprised the majority of each patient stability score. AutoTriage demonstrated an improvement in the sensitivity of existing stability prediction tools, while considering patient safety upon discharge. The relative contributions of risk factors indicated that time-series trends and measurement interdependencies are most important to stability prediction. Our results motivate the application of multidimensional analysis to other clinical problems and highlight the importance of risk factor trends and interdependencies in outcome prediction.

Novel wavelet neural network algorithm for continuous and noninvasive dynamic estimation of blood pressure from photoplethysmography

This paper proposes a novel wavelet neural network algorithm for the continuous and noninvasive dynamic estimation of blood pressure (BP). Unlike prior algorithms, the proposed algorithm capitalizes on the correlation between photoplethysmography (PPG) and BP. Complete BP waveforms are reconstructed based on PPG signals to extract systolic blood pressure (SBP) and diastolic blood pressure (DBP). To improve the robustness, Daubechies wavelet is implemented as the hidden layer node function for the neural network. An optimized neural network structure is proposed to reduce the computational complexity. Further, this paper investigates an inhomogeneous resilient backpropagation (IRBP) algorithm to calculate the weight of hidden layer nodes. The IRBP improves the convergence speed and reconstruction accuracy. Multiparameter intelligent monitoring in Intensive Care (MIMIC) databases, which contain a variety of physiological parameters captured from patient monitors, are used to validate this algorithm. The standard deviation $\sigma$ between reconstructed and actual BP signals is 4.4797 mmHg, which satisfies the American National Standards of the Association for the Advancement of Medical Instrumentation. The reconstructed BP waveform can be used to extract the SBP and DBP, whose standard deviations $\sigma$ are 2.91 mmHg and 2.41 mmHg respectively.