To help patients get accurate and timely detection of heart arrhythmias by biodata and AI

Wearable single lead ECG sensor for long-term heart monitoring & engine for heart disease detection in time for early intervention


iSoma developed a compact wearable ECG sensor, iRealcare, and an AI engine that generates expert-level reports based on the collected ECG data. The ECG sensor could be worn for two weeks. It continually sends ECG data to a cloud with an AI engine, which generates a medical report. The report is sent to medical practitioners in real time to assist with the timely diagnosis of multiple forms of arrhythmias.

Early diagnostics will help patients to get treatment in time and prevent serious diseases.


Electrocardiogram (ECG) Monitoring Module

A Wearable ECG Monitor for Deep Learning Based Real-Time Cardiovascular Disease Detection

Cardiovascular disease has become one of the significant threats endangering human life and health. The early prevention and treatment of cardiovascular disease make a difference between life and death. In the 21st century, mobile monitoring technology and information transmission technology have made significant progress. Hence, Electrocardiogram (ECG) monitoring has been transformed into remote cardiac monitoring by Holter surveillance. Holter are commonly used portable ECG devices nowadays using multi-channel ECG sensors, which brings a great deal of discomfort and inconvenience to the people who carry them. In order to reduce the burden on the wearer and dynamically monitor the patient’s heart status for a long time, various smaller-size light-weight wearable ECG patches have been developed. However, existing wearable ECG patches still have many drawbacks, such as short operation time and high price. In this work, we develop a new wireless ECG patch, and design an ECG IoT artificial intelligence (AI) diagnosis system for detecting cardiovascular diseases in real-time. The developed ECG system improves the data collection accuracy and increases the disease detection probability using deep learning algorithms.

As shown in Figure 1, the proposed system consists of three components: ECG sensing, IoT cloud, and client. An ECG sensor can generate an electrocardiogram by collecting electrical signals generated by each heartbeat of a human body. Figure 2 illustrates three main units of a sensor: a sensing unit, a control unit, and a wireless transmit unit.

  • In the sensing unit, the sensor collects ECG data and then transmits the data to the control unit.

  • The signals obtained in the control unit are sampled and processed at a sampling frequency of 250Hz. The sampled electrical signals are then converted to digital signals and ready to be sent to the wireless transmit unit. The 24-bit analog-to-digital converter (ADC) is used in our system implementation.

  • The wireless transmit unit interacts with the mobile device on the client side via Bluetooth.

ECG Signal

ECG is an objective indicator of the occurrence, transmission, and recovery of cardiac arousal. ECG provides essential reference information to the primary function of the heart and its case study. It can be used for all kinds of examinations on heart rate disorders or heart-related disorders, but also a vital basis to determine cardiovascular disease. As shown in Figure 6, there are three waveforms, two periods and two typical segments of ECG.

The three waveforms are (1) P waves produced by atrial agitation, the average p-wave time length is 0:12 second, and the average peak is 0:25mv.

When the atrium expands, the P wave can exhibit a higher tip or double-peak waves when the atria are abnormal. (2) The QRS wave group is the reaction of the point of the left and right ventricle. The QRS wave group represents the ventricular depolarization, and the agitation time length of normal duration should be less than 0:11 seconds. When the conduction block, ventricular enlargement or hypertrophy of the left and right bundle of the heart appears, the QRS wave group would be enlarged, deformed and prolonged. (3) T wave represents the potential of the recovery after ventricular agitation. The change of T wave is affected by a variety of factors, and normal T wave should be in the same direction as the QRS main wave.

ECG Monitor - IREALCARE 2.0

The overall appearance of IREALCARE2.0 is shown in Fig. 2(a) and (b). It is very compact and convenient. The size is 8 cm (length) 3.1 cm (width) 0.8 cm (thickness), and the weight is only 13 grams, therefore, it is very easy to wear for a long time without any discomfort. Fig. 2 (c) and (d) show the structure of IREALCARE 2.0. The core component of IREALCARE 2.0 is the printed circuit board (see Fig2 (e)). The ECG measurement chip and Bluetooth module are integrated on the PCB. ADS1291 is selected as the ECG measurement chip, which is a low power, 2-channel, 24-bit analog front-end for biopotential measurement. It integrates low-noise programmable gain amplifier (PGA) and high-resolution analog-to-digital converter (ADC). The function is to measure and amplify the ECG signal, and then convert the analog signal into a digital signal. Due to small size, low power consumption, high integration, and excellent performance, it has been widely used in medical measurement instruments (such as ECG measurement), signal acquisition and so on. The Bluetooth chip uses the nRF52832, it is a powerful, highly flexible, low-power multi-protocol Bluetooth chip. Some parameters of nRF52832 and other four Bluetooth chips are shown in Table I. Compared with other chips, nrf52832 has more memory, built-in 512KB Flash+64KB RAM, and has stronger computing power and lower power consumption.

IREALCARE2.0 can monitor the human’s heart for a long time, which can be worn on the chest. The ADS1291 chip is used to collect ECG signals, filter and amplify the ECG signal, and then convert analog signal into a digital signal. Finally, it is transmitted to the cloud server by Bluetooth and subsequent processing (shown as Fig. 3). And in the cloud server, the ECG signals can be analyzed and classified by our algorithms, and the results are transferred to user, family, and clinician.

Reference

1. L. Meng, K. Ge, Y. Song, D. Yang, and Z. Lin, Long-term wearable electrocardiogram signal monitoring and analysis based on convolutional neural network, the IEEE Transactions on Instrumentation & Measurement, Vol. 70, April 2021, DOI:10.1109/TIM.2021.3072144

2. Z. Chen, Z. Lin, P. Wang, and M. Ding, Negative-ResNet: noisy ambulatory electrocardiogram signal classification scheme, Neural Computing and Applications, Vol. 33, Issue 14, July 2021, pp. 8857-8869

3. P, Wang, Z. Lin, Z. Chen, X. Yan, M. Ding, A Wearable ECG Monitor for Deep Learning-Based Real-Time Cardiovascular Disease Detection, https://arxiv.org/abs/2201.10083

4. X. Yan, Z. Lin, and P. Wang, “Wireless electrocardiograph monitoring based on wavelet convolutional neural network,” in 2020 IEEE Wireless Communications and Networking Conference Workshops (WCNCW), 2020, pp. 1–6

5. X. Yan, D. Yang, Z. Lin, B. Vucetic. "Significant Low-dimensional Spectral-temporal Features for Seizure Detection". IEEE Transactions on Neural Systems & Rehabilitation Engineering 2022 Mar 4; PP. doi: 10.1109/TNSRE.2022.3156931.