A new machine learning method for predicting systolic and diastolic blood pressure using clinical characteristics

Healthcare Analytics, Journal Year: 2023, Volume and Issue: 4, P. 100219 - 100219

Published: June 28, 2023

Hypertension describes elevated blood pressure, which significantly impacts cardiovascular diseases. Typically, a sphygmomanometer, cuff-like device, is used to measure patient's pressure. However, new techniques such as phonocardiogram (PPG) and electrocardiogram (ECG) based on cuff signals have been developed. Still, they require complex expensive multiple sensors. A machine learning-based method has proposed predict both systolic diastolic pressure overcome this issue. The model considers various clinical characteristics gender, sugar cholesterol levels, smoking status, age, alcohol use, weight, history of heart disease. physical activity level metric evaluate the trained dataset 50,000 readings available Kaggle. Four learning techniques, including K-Nearest Neighbors (KNN), logistic regression, decision tree, random forest, were tested with different training, validation, testing ratios enhance model's accuracy. algorithm's performance was evaluated using accuracy, recall, precision, F1 scores. Random forest found highest accuracy

Language: Английский

---

Wearable blood pressure sensors for cardiovascular monitoring and machine learning algorithms for blood pressure estimation

Nature Reviews Cardiology, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 18, 2025

Language: Английский

---

PPG2ABP: Translating Photoplethysmogram (PPG) Signals to Arterial Blood Pressure (ABP) Waveforms

Bioengineering, Journal Year: 2022, Volume and Issue: 9(11), P. 692 - 692

Published: Nov. 15, 2022

Cardiovascular diseases are one of the most severe causes mortality, annually taking a heavy toll on lives worldwide. Continuous monitoring blood pressure seems to be viable option, but this demands an invasive process, introducing several layers complexities and reliability concerns due non-invasive techniques not being accurate. This motivates us develop method estimate continuous arterial (ABP) waveform through approach using Photoplethysmogram (PPG) signals. We explore advantage deep learning, as it would free from sticking ideally shaped PPG signals only by making handcrafted feature computation irrelevant, which is shortcoming existing approaches. Thus, we present PPG2ABP, two-stage cascaded learning-based that manages ABP input signal with mean absolute error 4.604 mmHg, preserving shape, magnitude, phase in unison. However, more astounding success PPG2ABP turns out computed values Diastolic Blood Pressure (DBP), Mean Arterial (MAP), Systolic (SBP) estimated outperform works under metrics (mean 3.449 ± 6.147 2.310 4.437 5.727 9.162 respectively), despite explicitly trained do so. Notably, both for DBP MAP, achieve Grade A BHS (British Hypertension Society) Standard satisfy AAMI (Association Advancement Medical Instrumentation) standard.

Language: Английский

---

A Shallow U-Net Architecture for Reliably Predicting Blood Pressure (BP) from Photoplethysmogram (PPG) and Electrocardiogram (ECG) Signals

Sensors, Journal Year: 2022, Volume and Issue: 22(3), P. 919 - 919

Published: Jan. 25, 2022

Cardiovascular diseases are the most common causes of death around world. To detect and treat heart-related diseases, continuous blood pressure (BP) monitoring along with many other parameters required. Several invasive non-invasive methods have been developed for this purpose. Most existing used in hospitals BP invasive. On contrary, cuff-based methods, which can predict systolic (SBP) diastolic (DBP), cannot be monitoring. studies attempted to from non-invasively collectible signals such as photoplethysmograms (PPG) electrocardiograms (ECG), In study, we explored applicability autoencoders predicting PPG ECG signals. The investigation was carried out on 12,000 instances 942 patients MIMIC-II dataset, it found that a very shallow, one-dimensional autoencoder extract relevant features SBP DBP state-of-the-art performance large dataset. An independent test set portion dataset provided mean absolute error (MAE) 2.333 0.713 DBP, respectively. an external 40 subjects, model trained MAE 2.728 1.166 For both cases, results met British Hypertension Society (BHS) Grade A surpassed current literature.

Language: Английский

---

KD-Informer: A Cuff-Less Continuous Blood Pressure Waveform Estimation Approach Based on Single Photoplethysmography

IEEE Journal of Biomedical and Health Informatics, Journal Year: 2022, Volume and Issue: 27(5), P. 2219 - 2230

Published: June 14, 2022

Ambulatory blood pressure (BP) monitoring plays a critical role in the early prevention and diagnosis of cardiovascular diseases. However, cuff-based inflatable devices cannot be used for continuous BP monitoring, while pulse transit time or multi-parameter-based methods require more bioelectrodes to acquire electrocardiogram signals. Thus, estimating waveforms only based on photoplethysmography (PPG) signals has essential clinical values. Nevertheless, extracting useful features from raw PPG fine-grained waveform estimation is challenging due physiological variation noise interference. For single analysis utilizing deep learning methods, previous works depend mainly stacked convolution operation, which ignores underlying complementary time-dependent information. this work presents novel Transformer-based method with knowledge distillation (KD-Informer) estimation. Meanwhile, we integrate prior information patterns, selected by backward elimination algorithm, into transfer branch KD-Informer. With these strategies, model can effectively capture discriminative through lightweight architecture during process. Then, further adopt an effective technique demonstrate excellent generalization capability proposed using two independent multicenter datasets. Specifically, first fine-tuned KD-Informer large high-quality dataset (Mindray dataset) then transferred pre-trained target domain (MIMIC dataset). The experimental test results MIMIC showed that exhibited error 0.02 ± 5.93 mmHg systolic (SBP) 0.01 3.87 diastolic (DBP), complied association advancement medical instrumentation (AAMI) standard. These high reliability elegant robustness measure waveforms.

Language: Английский

---

Emerging sensing and modeling technologies for wearable and cuffless blood pressure monitoring

npj Digital Medicine, Journal Year: 2023, Volume and Issue: 6(1)

Published: May 22, 2023

Abstract Cardiovascular diseases (CVDs) are a leading cause of death worldwide. For early diagnosis, intervention and management CVDs, it is highly desirable to frequently monitor blood pressure (BP), vital sign closely related during people’s daily life, including sleep time. Towards this end, wearable cuffless BP extraction methods have been extensively researched in recent years as part the mobile healthcare initiative. This review focuses on enabling technologies for monitoring platforms, covering both emerging flexible sensor designs algorithms. Based signal type, sensing devices classified into electrical, optical, mechanical sensors, state-of-the-art material choices, fabrication methods, performances each type briefly reviewed. In model review, contemporary algorithmic estimation beat-to-beat measurements continuous waveform introduced. Mainstream approaches, such pulse transit time-based analytical models machine learning compared terms their input modalities, features, implementation algorithms, performances. The sheds light interdisciplinary research opportunities combine latest innovations processing fields achieve new generation measurement with improved wearability, reliability, accuracy.

Language: Английский

---

Hybrid modeling on reconstitution of continuous arterial blood pressure using finger photoplethysmography

Biomedical Signal Processing and Control, Journal Year: 2023, Volume and Issue: 85, P. 104972 - 104972

Published: May 2, 2023

Language: Английский

---

A benchmark for machine-learning based non-invasive blood pressure estimation using photoplethysmogram

Scientific Data, Journal Year: 2023, Volume and Issue: 10(1)

Published: March 21, 2023

Abstract Blood Pressure (BP) is an important cardiovascular health indicator. BP usually monitored non-invasively with a cuff-based device, which can be bulky and inconvenient. Thus, continuous portable monitoring devices, such as those based on photoplethysmography (PPG) waveform, are desirable. In particular, Machine Learning (ML) estimation approaches have gained considerable attention they the potential to estimate intermittent or only single PPG measurement. Over last few years, many ML-based been proposed no agreement their modeling methodology. To ease model comparison, we designed benchmark four open datasets shared preprocessing, right validation strategy avoiding information shift leak, standard evaluation metrics. We also adapted Mean Absolute Scaled Error (MASE) improve interpretability of evaluation, especially across different datasets. The comes codes. showcase its effectiveness by comparing 11 three categories.

Language: Английский

---

PPG Signals-Based Blood-Pressure Estimation Using Grid Search in Hyperparameter Optimization of CNN–LSTM

Diagnostics, Journal Year: 2023, Volume and Issue: 13(15), P. 2566 - 2566

Published: Aug. 1, 2023

Researchers commonly use continuous noninvasive blood-pressure measurement (cNIBP) based on photoplethysmography (PPG) signals to monitor blood pressure conveniently. However, the performance of system still needs be improved. Accuracy and precision in measurements are critical factors diagnosing managing patients' health conditions. Therefore, we propose a convolutional long short-term memory neural network (CNN-LSTM) with grid search ability, which provides robust estimation by extracting meaningful information from PPG reducing complexity hyperparameter optimization proposed model. The multiparameter intelligent monitoring for intensive care III (MIMIC III) dataset obtained arterial-blood-pressure (ABP) signals. We 75,226 signal segments, 60,180 allocated training data, 12,030 validation set, 15,045 test data. During training, applied five-fold cross-validation grid-search method select best model determine optimal settings. optimized configuration CNN-LSTM layers consisted five layers, one (LSTM) layer, two fully connected estimation. This study successfully achieved good accuracy assessing both systolic (SBP) diastolic (DBP) calculating standard deviation (SD) mean absolute error (MAE), resulting values 7.89 ± 3.79 5.34 2.89 mmHg, respectively. provided satisfactory according standards set British Hypertension Society (BHS), Association Advancement Medical Instrumentation (AAMI), Institute Electrical Electronics Engineers (IEEE) devices.

Language: Английский

---

PulseDB: A large, cleaned dataset based on MIMIC-III and VitalDB for benchmarking cuff-less blood pressure estimation methods

Frontiers in Digital Health, Journal Year: 2023, Volume and Issue: 4

Published: Feb. 8, 2023

There has been a growing interest in developing cuff-less blood pressure (BP) estimation methods to enable continuous BP monitoring from electrocardiogram (ECG) and/or photoplethysmogram (PPG) signals. The majority of these have evaluated using publicly-available datasets, however, there exist significant discrepancies across studies with respect the size, number subjects, and applied pre-processing steps for data that is eventually used training testing models. Such differences make conducting performance comparison models largely unfair, mask generalization capability various methods. To fill this important gap, paper presents “PulseDB,” largest cleaned dataset date, benchmarking also fulfills requirements standardized protocols. PulseDB contains 1) 5,245,454 high-quality 10 -s segments ECG, PPG, arterial (ABP) waveforms 5,361 subjects retrieved MIMIC-III waveform database matched subset VitalDB database; 2) subjects’ identification demographic information, can be utilized as additional input features improve models, or evaluate generalizability unseen subjects; 3) positions characteristic points ECG/PPG signals, making directly usable deep learning minimal pre-processing. Additionally, dataset, we conduct first study provide insights about gap between calibration-based calibration-free approaches evaluating We expect PulseDB, user-friendly, large, comprehensive multi-functional reliable source evaluation

Language: Английский

---