Table 2.
Implementations of smart wearables in detection of CVDs.
| Ref# | Year | Disease(s) Targeted | Vital Signs Collected | Hardware Employed | Smart Model(s) Used | Training Dataset(s) | Results Metrics |
|---|---|---|---|---|---|---|---|
| [39] | 2010 | Atrial Fibrillation | Electrocardiogram | A wearable vest including dry foam ECG acquisition device A mobile phone (Nokia N85) |
Not Identified | PhysioNet MIT-BIH dataset | Sensitivity: 94.56% Positive Predictive Value: 99.22% |
| [40] | 2010 | Right Bundle Branch Block Beats Premature Ventricular Contraction Paced Beats Fusion of Paced and Normal Beats |
Electrocardiogram | Plug-In-Based GUI Platform: An Alive Bluetooth ECG heart monitor and Amoi E72 Microsoft Windows Mobile 5 Smartphone Machine-Learning-Based Platform: An Alive Bluetooth ECG heart monitor and an HTC Microsoft Windows Mobile 6 Smartphone |
Multilayer Perceptron | PhysioNet MIT-BIH dataset | Accuracy > 90% |
| [41] | 2010 | Sinus Tachycardia Sinus Bradycardia Cardiac Asystole Atrial Fibrillation Wide QRS Complex |
Electrocardiogram | A three-lead ECG device that contain two main parts: NCTU ECG Aquisition tool as the data acquisition (DAQ) unit and a wireless-transmission unit. Medi-Trace 200, Kendall are also used to read the ECG from the body |
Not Identified | Dataset collected at MUSE ECG system (GE health care, USA) in China Medical University (CMUH) database | Accuracy > 92% |
| [42] | 2011 | Premature Ventricular Contraction Atrial Premature Contraction |
Electrocardiogram Electroencephalogram Respiratory Rate Skin Temperature |
Wearable Sensor Node and it consists of seven modules: analog front-end circuits for four physiological signals, a radio communication module, a storage module, and MSP430F2618 as microcontroller unit (MCU) Smartphone: HTC HD2 with a 1 GHz CPU and 448 MB RAM (can be replaced with any android, Windows or IOS phone) |
Hidden Markov Model Layered Hidden Markov Model |
PhysioNet MIT-BIH dataset | Sensitivity: 99.72% Positive Predictive Value: 99.64% |
| [43] | 2011 | Congestive Heart Failure Malignant Ventricular Ectopy Ventricular Tachycardia |
Electrocardiogram | A wireless ECG sensor S3C6400 mobile phone HBE-ZigbeX motes as a wireless sensor network |
Multilayer Perceptron | PhysioNet MIT-BIH dataset | BIDMC Congestive Heart Failure: 100% Malignant Ventricular Ectopy: 90.9% Ventricular Tachyarrhythmia: 83.3% |
| [44] | 2015 | Atrial Fibrillation | Electrocardiogram | Rejiva ECG wearable sensor and a smartphone |
Support Vector Machines | PhysioNet MIT-BIH dataset | Specificity: 77.25% Sensitivity: 93.13% |
| [45] | 2016 | Atrial Fibrillation | Electrocardiogram Photoplethysmogram |
Samsung Simband wrist band smart watch | Elastic Net Logistic model | Private Data | Accuracy: 95% Sensitivity: 97% Specificity: 94% AUROC: 99% |
| [46] | 2016 | Myocardial Ischemia | Electrocardiogram | A smart cloth composed of four units: Smart cloth unit to measure physiological signal-ECG signal Signal control unit to control and memorize the status of the device by an ultra-low power MCU and SD card to save the signal data Signal sensing unit that has a motion tracking sensor module to capture the accelerometer signal Wireless connection unit to transmit the data A smartphone |
Neural Network | PhysioNet MIT-BIH dataset PhysioNet MIT-BIH Normal Sinus Rhythm dataset |
Accuracy > 76% |
| [47] | 2017 | Atrial Fibrillation | Electrocardiogram Photoplethysmogram |
Samsung Simband wrist band smart watch | Convolutional Neural Network Elastic Net Logistic model |
Private Data | Accuracy: 91.8% |
| [48] | 2017 | Heart Attack | Electrocardiogram Body Temperature |
Device composed of pulse sensor, a temperature sensor, an Arduino, and a Low Energy (LE) Bluetooth A smartphone |
Not Identified | Private Data | |
| [49] | 2017 | Ventricular Premature Complex Atrial Premature Complex Ventricular Fibrillation Atrial Fibrillation |
Electrocardiogram | Bio Clothing One, XYZ life BC1 | Artificial Neural Networks | PhysioNet American Heart Association database PhysioNet Creighton University Ventricular Tachyarrhythmia database PhysioNet MIT-BIH dataset PhysioNet MIT-BIH Noise Stress Test database |
Accuracy > 75% |
| [50] | 2017 | Atrial Fibrillation | Electrocardiogram | Wrist bracelet designed for the purpose: based on the ultra low power series Microcontroller STM32L471RG | Support Vector Machines | Private Data | Accuracy: 95% |
| [51] | 2017 | Atrial Fibrillation | Audio Signal in Radial Artery | The PAG monitoring device consists of four components audiogram sensor: Panasonic capacitive microphone analog-digital converter: Embedded in Atmega328P microprocessor: Atmega328P chip data storage unit A smartphone |
Convolutional Neural Network | Dataset collected at National Cheng Kung University Hospital (NCKUH), Tainan, Taiwan. | Accuracy: 98.92% |
| [52] | 2018 | Myocardial Infarction | Electrocardiogram | ECG sensor using AD8232 and Espressif ESP-32 Wi-Fi + BLE module | Convolutional Neural Network | PhysioNet PTB Diagnostic ECG Database | Accuracy: 84% |
| [53] | 2018 | Ventricular Arrhythmia Junctional Arrhythmia Supraventricular Arrhythmia Arrhythmias |
Electrocardiogram | a smart clothing consisting of cloth carrier, biosen sor platform, and smart terminals. In biosensor platform, ADI ECG analog front-end (ADAS1001) is used for obtaining the ECG signals, Microcontroller (STM32) is used to realize the data processing and a Bluetooth module is available for data transfer | Deep Neural Network with a Softmax Regression model | PhysioNet MIT-BIH dataset | Accuracy > 94% |
| [54] | 2018 | Hypertension | Heart Rate | A waist belt comprised of three kinds of sensors: three dry electrodes, a 3-axis accelerometer and two pressure sensors with different sensitivities | Logistic Regression Support Vector Machines |
Private Data | Accuracy: 93.33% |
| [55] | 2018 | Atrial Fibrillation | Electrocardiogram Photoplethysmogram |
Samsung gear device wearable device | Convolution–Recurrent Hybrid Model (CRNN) | Private Data | Accuracy > 98% |
| [56] | 2018 | Atrial Fibrillation | Electrocardiogram | A smart shirt equipped with ECG sensors A smartphone |
Dataset collected at the Dongsan Medical Center in South Korea | Accuracy: 98.2% | |
| [57] | 2018 | Ventricular Tachycardia Ventricular Bradycardia Premature Atrial Contractions Premature Ventricular Contractions |
Electrocardiogram | for ECG Sensing: ECG body sensor with analog conditioning circuit (AD8232), Microcontroller unit (MCU) (PIC12F1822), Bluetooth module (HC-06), and charging controller module for analysis and display: processing and displaying unit of that process the ECG signal and display it on thin film transistor (TFT) liquid crystal display (LCD) consisting of Rpi computer, Bluetooth module, TFT screen, and power supply |
Support Vector Machines | PhysioNet MIT-BIH dataset | Accuracy: 96.2% |
| [58] | 2019 | Myocardial Infarction Heart Failure Arrhythmias Fusion Beats Supraventricular Ectopic Beats Ventricular Ectopic Beats |
Electrocardiogram Heart Rate Respiratory Rate |
A patch with electronic circuit is built for the purpose and proposed in the article and an Android smartphone and a cloud server for data storage and further analysis | Convolutional Neural Network | PhysioNet PTB Diagnostic ECG Database St Petersburg INCART 12-lead Arrhythmia Database |
Accuracy: 98.7% |
| [59] | 2019 | Atrial Fibrillation | Electrocardiogram | A patch with electronic circuit is built for the purpose and proposed in the article and an Android smartphone and a cloud server for data storage and further analysis | Decision Tree | PhysioNet MIT-BIH dataset | Accuracy > 97.18% |
| [60] | 2019 | Atrial Fibrillation Atrial Flutter Ventricular Fibrillation |
Electrocardiogram | A wearable ECG sensing device and an Android smartphone and a cloud server for data storage and further analysis | Convolutional Neural Network | PhysioNet MIT-BIH dataset | Accuracy > 94% |
| [61] | 2019 | Atrial Fibrillation | Electrocardiogram | Smart vest equipped with two ECG sensing units | Long Short-Term Memory | PhysioNet dataset of the 2017 Computing in Cardiology Challenge | Sensitivity: 83.82% Specificity: 97.84% F1-score: 81.43% |
| [62] | 2019 | Supraventricular Ectopic Beats Ventricular Ectopic Beats |
Electrocardiogram | ECG sensing device with a smartphone or tablet | Long Short-Term Memory | PhysioNet MIT-BIH dataset | Accuracy > 79% |
| [63] | 2019 | Atrial Fibrillation | Heart Rate | Commercial HR Sensor | Long Short-Term Memory | PhysioNet Atrial Fibrillation Database (AFDB) | Accuracy: 98.51% |
| [64] | 2019 | Arrhythmias Congestive Heart Failure |
Electrocardiogram | One lead ECG sensor | Convolutional Neural Network | PhysioNet MIT-BIH dataset PhysioNet MIT-BIH Normal Sinus Rhythm database |
Accuracy: 93.75% |
| [65] | 2019 | Arrhythmias | Electrocardiogram | A device composed of a single-lead heart rate monitor front end AD8232 chip, Atmel’s ATmega128 as a microcontroller and a BLE module A smartphone is also used |
Support Vector Machines K-Nearest Neighbors Logistic Regression Random Forest Decision Tree Gradient Boosting Decision Tree |
PhysioNet MIT-BIH dataset | Accuracy > 77% |
| [66] | 2019 | Atrial Fibrillation | Photoplethysmogram | Wearable wristband device | Support Vector Machines | Private Data | Accuracy: 90% |
| [67] | 2020 | Atrial Bigeminy Atrial Fibrillation Atrial Flutter Ventricular Bigeminy Heart Block Ventricular Trigeminy Ventricular Flutter Ventricular Tachycardia Supraventricular Tachyarrhythmia Idioventricular Rhythm Paced Beats Nodal (A-V Junctional) Rhythm |
Electrocardiogram | SparkFun Single Lead Heart Rate Monitor AD8232 as the data acquisition device Smartphone as a gateway to the server |
Convolutional Neural Network | PhysioNet MIT-BIH dataset | Accuracy: 94:13% |
| [68] | 2020 | Atrial Fibrillation | Electrocardiogram Photoplethysmogram |
Amazfit Healthband 1S for ECG and PPG sensing smartphone for data reception and analysis |
Convolutional Neural Network | Dataset collected at Peking University First Hospital | Sensitivity: 80.00% Specificity: 96.81% Accuracy: 90.52% |
| [69] | 2020 | Left Bundle Branch Block Beats Right Bundle Branch Block Beats Atrial Premature Contraction Ventricular Premature Contraction Paced Beats Ventricular Escape Beats |
Electrocardiogram | A sensing device composed from a single lead heart rate monitor AD8232 and interfaced with NodeMCU development board having ESP8266 microcontroller capable of connecting to internet via WiFi Smartphone for the analysis of the data |
Convolutional Neural Network | PhysioNet MIT-BIH dataset | Accuracy > 90% |
| [70] | 2020 | Cardiovascular Risk | Electrocardiogram Electroencephalogram Electromyogram Heart Rate Blood Pressure Respiratory Rate Blood Sugar Level Oxygen Saturation Level Cholesterol Levels |
Wearable medical sensors and a wearable smart watch | Convolutional Neural Network | UCI Cleveland Heart Diseases Dataset | Accuracy: 98.5% |
| [71] | 2020 | Atrial Fibrillation | Electrocardiogram Photoplethysmogram Photoplethysmogram Oxygen saturation Level Body Temperature |
The sensing device used is composed of three parts: AD8232r for ECG detection, ADS1115 analog-to-digital converter and SX1276 LoRa chip that transmits the data to the fog device The fog device: a low-cost raspberry pi system integrated with Intel Neural Compute Stick 2 (NCS 2) that is capable of handling deep learning algorithms |
Convolutional Neural Network | PhysioNet dataset of the 2017 Computing in Cardiology Challenge | Accuracy: 90% |
| [72] | 2020 | Cardiovascular Risk | Electrocardiogram Blood Pressure |
An ECG sensing device built with AD8232 unit A smart watch raspberry pi with SX1272 unit to transmit the data for LoRa gateway |
Convolutional Neural Network | UCI Cleveland Heart Diseases Dataset | Accuracy: 98.2% |
| [73] | 2020 | Aortic Stenosis Mitral Insufficiency Mitral Stenosis Tricuspid Regurgitation |
Electrocardiogram Photoplethysmogram Gyrocardiography Seismocardiogram |
Shimmer 3 from Shimmer Sensing for ECG detection A three-axis MEMS accelerometer: (Kionix KXRB5-2042, Kionix, Inc.) to measure the SCG signal A three-axis MEMS gyroscope (Invensense MPU9150, Invensense, Inc.) to record the GCG signal An ear-lobe photoplethysmography (PPG) sensor |
Decision Tree Random Forest Neural Network |
Dataset collected at Columbia University Medical Center (CUMC) | Accuracy > 90% |
| [74] | 2020 | Left Bundle Branch Block Beats Right Bundle Branch Block Beats Atrial Escape Beats Nodal (Junctional) Escape Beats Atrial Premature Beats Aberrated Atrial Premature Beats Nodal Premature Beats Supraventricular Premature Beats Premature Ventricular Contractions Ventricular Escape Beats Fusion of Ventricular and Normal Beats Paced Beats Fusion of Paced and Normal Beats |
Electrocardiogram | A sensing device composed of AD8232 single-lead three-electrode ECG Heart Rate monitor and a ESP8266 Wi-Fi module used to provide wireless data transmission access to the Arduino Nano and is used to connect it to the cloud | Convolutional Neural Network | PhysioNet MIT-BIH dataset | Accuracy: 99.625% Sensitivity: 97.736% Specificity: 99.713% Precision: 97.835% |
| [75] | 2020 | Ventricular Ectopic Beats Arrhythmias |
Electrocardiogram | Sensing device composed of Raspberry Pi for processing, ADS1115 as Analog to Digital Converter and AD8232 as ECG sensor | Convolutional Neural Network | PhysioNet MIT-BIH dataset | Accuracy: 95.76% |
| [76] | 2020 | Premature Atrial Contractions Premature Ventricular Contractions Atrial Fibrillation |
Electrocardiogram Photoplethysmogram |
7-lead Holter monitor (Rozinn RZ153+ Series, Rozinn Electronics Inc., Glendale, NY, USA) Smartwatch (Simband 2, Samsung Digital Health, San Jose, CA, USA) |
Random Forest Support Vector Machines |
Dataset collected at the ambulatory cardiovascular clinic at the University of Massachusetts Medical Center (UMMC) | Best Model Accuracy: 94% |
| [77] | 2020 | Arrhythmias | Electrocardiogram | Sensing device built using Raspberry Pi 3 model B+ and two ECG sensors AD8232 with a pulse sensor and an analog digital converter ADS1015 | Support Vector Machines Naïve Bayes Artificial Neural Networks |
PhysioNet MIT-BIH dataset | Best Model Accuracy: 97.8% |
| [78] | 2020 | Atrial Fibrillation | Electrocardiogram | the wearable system is composed to work on a prototype developed by Medicaltech srl (Rovereto, Italy) | A Custom model based on Thresholding of Shannon Entropy values | PhysioNet MIT-BIH dataset | Sensitivity: 99.2% Specificity: 97.3% |
| [79] | 2020 | Atrial Fibrillation | Electrocardiogram | The sensing device is composed of Raspberry pi 3, Arduino UNO, AD8232 single lead ECG sensor, HC-05 Bluetooth, biomedical sensor pad and battery | Long Short-Term Memory | PhysioNet MIT-BIH dataset | Accuracy: 97.57% |
| [80] | 2020 | Atrial Escape Beats Junctional Escape Beats Left Bundle Branch Block Beats Right Bundle Branch Block Beats Atrial Premature Beats Aberrated Atrial Premature Beats Junctional Premature Beats Supraventricular Premature Beats Premature Ventricular Contractions Ventricular Escape Beats Fusion of Ventricular and Normal Beats Paced Beats Fusion of Paced and Normal Beats |
Electrocardiogram | Moto 360 NanoPi Neo Plus2 Raspberry Pi Zero |
Long Short-Term Memory | PhysioNet MIT-BIH dataset | Accuracy > 98.6 % |
| [81] | 2020 | Supraventricular Arrhythmia Atrial Fibrillation Arrhythmias |
Electrocardiogram | A wearable sensing device composed of AD8232 as an ECG sensor, MCP3008 ias an ADC and Raspberry Pi as a computing unit | Support Vector Machines | UCI Cleveland Heart Diseases Dataset | Accuracy: 72.41% |
| [82] | 2020 | Arrhythmias | Electrocardiogram Body Temperature Heart Rate Blood Oxygen Level |
A sensing device composed of: Temperature sensor: MLX90614 Heart rate and blood oxygen sensors: MAX30100 ECG sensor: AD8232 Inter-Integrated Circuit (I2C) communication protocol Microcontroller: Arduino UNO Wireless transmission: Wi-Fi chip ESP8266 A smartphone |
Long Short-Term Memory Convolutional Neural Network |
PhysioNet MIT-BIH dataset | Accuracy: 99.05% |
| [83] | 2020 | Premature Ventricular Contraction | Electrocardiogram | A wireless 3-lead ECG sensor from Shimmer Sensing | Support Vector Machines | PhysioNet MIT-BIH dataset | Sensitivity: 96.51% Predictive Value: 81.92% |
| [84] | 2020 | Atrial Fibrillation Syncope |
Electrocardiogram | A sensing device composed of: The SparkFun AD8232 ECG sensing unit Arduino Mega 2560 microcontroller Raspberry Pi 3 board ADXL345 triple-axis accelerometer HC-05 Bluetooth sensor A smartphone |
Long Short-Term Memory | PhysioNet MIT-BIH dataset | Accuracy: 97.61% |
| [85] | 2021 | Atrial Fibrillation | Pulse Plethysmogram | Wrist-type pulse wave monitor (type: Smart TCM-I, product by: Shanghai Asia & Pacific Computer Information System CO, Ltd, Shanghai, China) |
Time Synchronous Averaging | Private Data | Accuracy: 98.4% |
| [86] | 2021 | Cardiovascular Risk | Photoplethysmogram | Pulse rate sensor with ATmega32 microcontroller | Support Vector Machines Naïve Bayes Random Forest Decision Tree Logistic Regression Artificial Neural Networks Recurrent Neural Networks |
Dataset collected at Framingham University | Accuracy: 94.9% |
| [87] | 2021 | Ventricular Ectopic Beats Supraventricular Ectopic Beats |
Electrocardiogram | Ternary second-order delta modulator circuits | Support Vector Machines | PhysioNet MIT-BIH dataset | Accuracy > 98% |
| [88] | 2021 | Premature Atrial Contractions Premature Ventricular Contractions Atrial Fibrillation Ventricular Tachycardia Sinus Bradycardia Atrial Tachycardia |
Electrocardiogram | A custom-built ECG Signal acquisition circuit | Gramian Angular Fields (GAFs) Deep Residual Network (ResNet) |
PhysioNet MIT-BIH dataset LTAF database Simulated Data (Prosim2 Vital Sign Simulator) |
Accuracy: 98.1% Sensitivity: 97.6% Specificity: 99.7% F1 Score: 97.6% |
| [89] | 2021 | Arrhythmias Congestive Heart Failure |
Electrocardiogram | ARDUINO UNO ECG SENSOR AD8232 DISPOSABLE ECG ELECTRODES |
Support Vector Machines | PhysioNet dataset of the 2016 Computing in Cardiology Challenge | Accuracy: 98% |
| [90] | 2021 | Atrial Fibrillation | Electrocardiogram | A consumer-grade, single-lead heart belt (Suunto Movesense, Suunto, Vantaa, Finland) | Not Identified | Private Data | Accuracy 97.8% |
| [91] | 2021 | Atrial Fibrillation Atrial Flutter Supraventricular Tachycardia Ventricular Tachycardia |
Electrocardiogram | ECG247 Smart Heart Sensor | Not Identified | Private Data | Accuracy > 95% |
| [92] | 2021 | Heart Attack | Electrocardiogram Heart Rate Body Temperature Blood Pressure |
A device composed of ECG, heart rate, body temperature, and blood pressure sensors | Not Identified | Private Data | Accuracy: 83% |
| [93] | 2021 | Atrial Fibrillation Ventricular Bradycardia Ventricular Tachycardia Bundle Branch Block |
Electrocardiogram | HealthyPiV3 biosensors | Convolutional Neural Network | PhysioNet MIT-BIH dataset PhysioNet PAF Prediction Challenge Database for AF records PhysioNet PTB Diagnostic ECG Database PhysioNet dataset of the of 2015 bradycardia Challenge PhysioNet Fantasia Database and PAF Prediction Challenge Database for healthy signals |
Accuracy > 98.75% |
| [94] | 2021 | Heart Attack | Electrocardiogram | AD8232 ECG sensor | Sequential Covering Algorithm | PhysioNet PTB-XL dataset | F1 Score: 87.8% |
| [95] | 2021 | Heart Attack | Electrocardiogram Body Temperature Activity Parameters Oxygen Saturation Level |
Composed of different sensors to collect different vital signs which are: LM35, MPU 6050, MAX30100 and AD8232 respectively | Support Vector Machines Linear Regression K-Nearest Neighbors Naïve Bayes |
Private Data | Accuracy: 80% |
| [96] | 2021 | Ventricular Premature Beats Supraventricular Premature Beats Atrial Fibrillation |
Electrocardiogram | IREALCARE2.0 Wearable ECG Sensor | Time-Span Convolutional Neural Network Recurrent Neural Networks |
Private Data | F1 Score: 86.5% Precision: 87.7% Recall: 86.8% |
| [97] | 2021 | Cardiovascular Risk | Electrocardiogram Oxygen Saturation Level |
Composed of AD8232 (ECG sensor) and MAX30102 (SPO2 sensor) | Convolutional Neural Network Convolutional Neural Network |
PhysioNet MIT-BIH dataset | Shallow CNN Accuracy: 96.06% Deep CNN Accuracy: 98.47% |
| [98] | 2021 | Heart Failure Hypertension Atrial Fibrillation Peripheral Artery Disease Myocardial Contraction |
Heart Rate Activity Parameters |
GENEActiv and Activinsights Band (Activinsights Ltd., Kimbolton, UK) |
Not Identified | To be collected | To be provided |
| [99] | 2021 | Atrial Fibrillation | Heart Rate Respiratory Rate |
BioHarness 3.0 by Zephyr | Support Vector Machines | PhysioNet MIT-BIH dataset | Sensitivity: 78% Specificity: 66% |
| [100] | 2021 | Atrial Fibrillation Bigeminy Arrhythmias |
Electrocardiogram | AD8232 | Decision Tree | Private Data | Accuracy > 95% |
| [101] | 2021 | Atrial Fibrillation Atrial flutter Left Bundle Branch Block Beats Wolff-Parkinson-White Syndrome Atrial Premature Contraction Premature Ventricular Contraction |
Electrocardiogram | A smart vest equipped with AD8232 ECG Sensor | Shallow Wavelet Scattering Network (ScatNet) | PhysioNet MIT-BIH dataset | Accuracy > 96% |
| [102] | 2021 | Tachycardia | Heart Rate Respiratory Rate Blood Oxygen Level |
Medical-grade wearable embedded system (SensEcho, Beijing SensEcho Science & Technology Co Ltd) | Long Short-Term Memory | Medical Information Mart for Intensive Care III (MIMIC-III) | Up to 80% accuracy 2 h before onset of Tachycardia |
| [103] | 2021 | Atrial Fibrillation | Photoplethysmogram | Samsung Galaxy Active 2 Watch | Convolutional Neural Network | Private Data | Accuracy 91.6% Specificity 93.0% Sensitivity 90.8% |
| [104] | 2021 | Arrhythmias | Electrocardiogram | A chest sticker that is composed from BMD101 ECG sensing device with YJ33 power supply, BQ24072 as a power source and JDY-30 as a Bluetooth module | Convolutional Neural Network | PhysioNet MIT-BIH dataset | Accuracy: 99.83% |
| [105] | 2022 | Supraventricular Ectopic Beats Ventricular Ectopic Beats Fusion Beats |
Electrocardiogram | Polar H10 | Decision Tree Gradient Boosting k-Nearest Neighbors Multilayer Perceptron Random Forest Support Vector Machines |
PhysioNet MIT-BIH dataset | Best Model Accuracy: 99.67% |
| [106] | 2022 | Supraventricular Ectopic Beats Ventricular Ectopic Beats Fusion Beats |
Electrocardiogram | Polar H10 | Decision Tree Gradient Boosting k-Nearest Neighbors Multilayer Perceptron Random Forest Support Vector Machines |
PhysioNet MIT-BIH dataset | Best Model Accuracy: 99% |
| [107] | 2022 | Heart Failure Reduced Ejection Fraction |
Electrocardiogram | Galaxy Watch Active & AppleWatch 6 | Convolutional Neural Network | Private Data | Area Under Curve 93.4% |
| [108] | 2022 | Atrial Fibrillation | Photoplethysmogram Electrocardiogram |
Samsung GalaxyWatch Active 2 Chest ECG Patch |
Hybrid Decision Model | Private Data | Average: 67.8% |
| [109] | 2022 | Atrial Fibrillation | Photoplethysmogram | Custom-built device that contains the PPG sensor MAX30102 | Convolutional Neural Network | Data obtained from Kaunas University of Technology | F1-score: 94% |
| [110] | 2022 | Atrial Fibrillation | Electrocardiogram | Firstbeat Bodyguard 2, Firstbeat Technologies | Not Identified | Private Data | Accuracy 98.7% Sensitivity 99.6%, Specificity 98.0% |
| [111] | 2022 | Supraventricular Ectopic Beats Ventricular Ectopic Beats |
Electrocardiogram | Custom-built device that contains the ECG AFE sensor | Artificial Neural Networks Decision Tree K-Nearest Neighbors |
PhysioNet MIT-BIH dataset | Accuracy: 98.7% |
| [112] | 2022 | Atrial Fibrillation | Photoplethysmogram | Apple Watch | Gradient Boosting Decision Tree | Private Data | Accuracy: 94.16% |
| [113] | 2022 | Congestive Heart Failure Atrial Fibrillation |
Electrocardiogram | AD8232 sensor | Random Forest | PhysioNet MIT-BIH dataset | Accuracy: 85% |
| [114] | 2022 | Cardiovascular Risk | Photoplethysmogram Body Temperature Activity Parameters |
Custom-built device with Pulse Sensor, DS18B20 temperature sensor and ADXL 1335 as accelerometer sensor | Naïve Bayes Decision Tree K-Nearest Neighbors Support Vector Machines |
Kaggle Human Gait Dataset Kaggle Heart Disease Prediction Dataset |
Accuracy: 82% |
| [115] | 2022 | Cardiovascular Risk | Heart Rate Respiratory Rate Blood Oxygen Level |
Not identified (WBAN) | Enhanced version of Recurrent Neural Network named ERNN | Private Data | Accuracy: 96% |
| [116] | 2022 | Cardiovascular Risk | Electrocardiogram Electroencephalogram Body Temperature Blood Oxygen Level Respiratory Rate Blood Sugar Level |
A custom-built device equipped with electrocardiogram sensor, electroencephalogram sensor, an electro-mammography sensor, an oxygen level sensor, a temperature sensor, a respiration rate sensor, and a glucose level sensor | Long Short-Term Memory | UCI Cardiac Arrhythmia Dataset | Average Positive Predictive Value: 96.77% Average Negative Predictive Value: 95.12% Average Sensitivity: 95.30% |
| [117] | 2022 | ST Elevation Myocardial Infarction (STEMI) | Electrocardiogram Motion Data |
Custom-built device with 3-axis accelerometer (ADXL355), 3-axis gyroscope (LSM6DS3) and single-lead ECG sensors |
Logistic Regression | Private Data | Sensitivity: 73.9% Specificity: 85.7% |
| [118] | 2022 | Cardiovascular Risk | Electrocardiogram Motion Data |
A custom-built device with accelerometers, Galvanic Skin Response (GSR) and electrocardiograms (ECG) sensors |
Mixed Kernel Based Extreme Learning Machine (MKELM) | Private Data | Accuracy: 99.5% |
| [119] | 2022 | Cardiovascular Risk | Heart Rate | Wrist Strap & Rohm BH1790GLC-EVK-001 Development board BH1790GLC Optical heart rate sensor | Convolutional Neural Network | Simulated Data | F1-Score: Up to 99% |
| [120] | 2022 | Myocardial Infarction Dilated Cardiomyopathy Hypertension |
Pulse Plethysmogram | PTN-104 PPG sensor | Support Vector Machines K-Nearest Neighbors Decision Tree |
Private Data | Accuracy: 98.4% Sensitivity: 96.7% Specificity: 99.6% |
| [121] | 2022 | Cardiovascular Risk | Heart Rate Blood Sugar Level |
Heart rate sensor by Sunrom Electronics Glucose monitor by Medtonic |
Naïve Bayes K-Nearest Neighbors Support Vector Machines Random Forest Artificial Neural Networks |
Private Data | Accuracy: 97.32% Recall: 97.58% Precision: 97.16% F1-Measure: 97.37% Specificity: 96.87% G-Mean: 97.22% |
| [122] | 2022 | Cardiovascular Risk | Electrocardiogram | A custom-built device composed of ECG sensor (AD8232) and other components | Random Forest | UCI Cleveland Heart Diseases Dataset | Accuracy: 88% |
| [123] | 2022 | Cardiovascular Risk | Heart Rate Oxygen Saturation Level Systolic Pressure Diastolic Pressure |
Custom-built soft transducer equipped with MAX30100 SpO2 and HR monitor sensor | Long Short-Term Memory | Kaggle dataset (Not Specified) | Accuracy > 93% |
| [124] | 2022 | Cardiovascular Risk | Electrocardiogram Blood Pressure Pulse Plethysmogram Body Temperature |
Custom-built device equipped with ECG sensor, TMP117 temperature sensor, Honeywell’s 26 PC SMT blood pressure sensor, and a pulse oximeter | Recurrent Neural Networks | UCI Cleveland Heart Diseases Dataset | Accuracy: 99.15% Precision: 98.06% Recall: 98.95% Specificity: 96.32% F1-Score: 99.02% |
| [125] | 2022 | Congenital Heart Disease | Electrocardiogram Seismocardiogram |
Custom-built chest wearable sensor equipped with ECG sensor (ADS1291; Texas Instruments, Dallas, TX) and seismocardiogram sensor (ADXL355; Analog Devices, Norwood, MA) | Ridge Regression | Private Data | - |