From Wearable Device to OpenEMR: 5G Edge Centered Telemedicine

and Decision Support System

Ying Wang

1,2

, Patricia Tran

1,2

and Janusz Wojtusiak

1,2

School of Systems and Enterprises, Stevens Institute of Technology, Hoboken, NJ, U.S.A.

Health Informatics Department, George Mason University, Fairfax, U.S.A.

Keywords:

IoT, 5G, Wearable Communication, Mobile Edge Computing, Cloud/Edge, Artiﬁcial Intelligence, OpenEMR,

Atrial Fibrillation.

Abstract:

The Internet of Things (IoT) is developing rapidly, with applications across various ﬁelds and industries. In

healthcare, wearable devices and the Internet of Medical Things (IoMT) have tremendous potential for im-

provements in the quality of telemedicine and producing medical insights and discoveries. Massive Machine

Type of Communication (mMTC) in 5G further reduces latency and enhances connectivity in supporting

wearables and IoMT, which provides a promising infrastructure for telemedicine. Although cloud computing

reduced the computation and storage load on wearable devices signiﬁcantly, the massive amounts of data pro-

duced by wearable devices and IoMT introduce challenges for latency and storage in the cloud. Additionally,

applications will need to navigate the regulation and compliance laws related to handling sensitive and private

health data, adding complexity to the accessibility and distribution of such innovations. This study ﬁrst exam-

ined the current frameworks for wearable devices in 5G telemedicine implementation and discussed existing

challenges. We then proposed a multi-layer 5G mobile edge computing (MEC) centered telemedicine design

that dynamically integrates wearable devices with OpenEMR electronic health records system. The multi-

layer design includes the IoT layer, MEC layer, Network layer, and Application layer. Near-real-time artiﬁcial

intelligence (AI) components and electronic health record (EHR) instances are automatically deployed to and

removed from the MEC layer to keep cloud computing capabilities closest to the infrastructure edge when a

user is associating and disassociating with a 5G bases station, respectively. Lastly, we demonstrate a proof

of concept by designing and implementing a system for detecting atrial ﬁbrillation (Aﬁb) over the design we

proposed. Aﬁb detection has the character of predictable trending, random occurrence of adverse events, and

urgent care needed when happening. These characters requires a low latency, large range coverage and high

throughput infrastructure. The proposed approach provides a distributed solution addressing the requirements

for Aﬁb detection. This approach can be used for other applications in telemedicine beyond Aﬁb detection.

1 INTRODUCTION

The Internet of Medical Things (IoMT) improves

multiple aspects in healthcare, including asset man-

agement in hospitals, patients’ vitals remote monitor-

ing, treatment compliance monitoring, smarter med-

ication, assisted living, and telemedicine, etc. In

IoMT, various medical devices or sensors, smart-

phones, imaging devices, personal digital assistants,

and electronic health records (EHR) integrate and

act as core parts of the system (Latif et al., 2017).

At present, wearable biomedical/health devices are

developing rapidly, offering advantages such as the

continuity of medical services and real-time capture

of health data. Wearable devices consist of sen-

sors placed on the body to capture and monitor data.

Ranging from ﬁtness trackers and smartwatches to

augmented reality (AR)/virtual reality (VR) glasses,

wearable devices in healthcare collect data on a vari-

ety of measures such as heart rates, sleeping cycles,

locations, and steps, etc. (Haghi et al., 2017). Wear-

able devices offer easier access, mobility, and conve-

nience for users and medical personnel and have al-

ready demonstrated applications in fall identiﬁcation

and prevention, physical activity monitoring, sports

medicine, patient education, diabetes care manage-

ment, and more (Min Wu and Jake Luo, 2020). The

emerging and integration of artiﬁcial intelligence (AI)

and machine learning (ML), big data, and IoT has en-

hanced the degree of intelligence of wearable devices

Wang, Y., Tran, P. and Wojtusiak, J.

From Wearable Device to OpenEMR: 5G Edge Centered Telemedicine and Decision Support System.

DOI: 10.5220/0010837600003123

In Proceedings of the 15th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2022) - Volume 5: HEALTHINF, pages 491-498

ISBN: 978-989-758-552-4; ISSN: 2184-4305

491

(Zhang et al., 2020). Combined with 5G technology,

these technologies have the potential to revolutionize

the healthcare industry, facilitating exciting research

and development towards this direction in the ﬁeld of

telemedicine (Latif et al., 2017).

In current implementations of telemedicine over

the IoMT monitoring in the both Long Term Evolu-

tion (LTE) and Wi-Fi system, three different types of

operation exist. The ﬁrst category is wearable de-

vices to collect health data and connect to the cloud

in the data network. In the cloud, AI may take place

to provide diagnostic or predictive analysis for physi-

cians or patients, and decision-making support for

treatment or management (Zhang et al., 2020). The

second category is for IoMT devices to run in an ac-

cessory mode and tether to a local connected mobile

device for data storage and processing. An IoMT in

the accessory mode is connected to a mobile phone

through Wi-Fi or Bluetooth for raw data transmis-

sion, and most of the data analysis is processed on

the mobile phone. The limitation of this model is

the requirement of mobile phone presence within Wi-

Fi/Bluetooth coverage, which is usually within 10 me-

ters. The third category is for wearable devices to

perform in standalone mode. For IoT in standalone

mode, the connection to a base station is through cel-

lular technology directly, and local data analysis oc-

curs without any connection to the cloud. However,

computing and energy limitations continue to hinder

these devices’ ability to process data locally. Addi-

tionally, there are limitations in hardware cost, size,

and the number of devices able to connect to the base

station at a given time.

Wearable devices are limited by power and stor-

age constraints, hardware size, and computing capa-

bility and are subject to high hardware costs. Thus,

most current IoMT devices adopted the second and

third category model by sending their data to mobile

devices with more computation power or further con-

necting to the clouds for processing (Sun et al., 2018).

Existing technology such as LTE-Advanced and Wi-

Fi are gradually evolving to ﬁt the needs of wearable

communication in telemedicine. However, besides

the aforementioned limitations, telemedicine over the

IoT will generate an unprecedented amount of data

requiring transmission, analysis, and storage and face

challenges such as security, latency, and connectivity

under the current LTE infrastructure.

Furthermore, IoMT devices handles sensitive and

private data, but the current regulations and policy

of data generated by wearable data are not sufﬁcient.

Unlike traditional health or medical data required to

comply with HIPAA regulations, HIPAA protection

does not extend to wearables and Apps. Integrat-

ing wearable devices with regulated EHR systems

like OpenEMR offers advantages of HIPPA protec-

tion and the creation of new applications. The cy-

ber security of healthcare data is more stringent than

other areas, and health data is frequently the target

of cybercriminals. Despite several digital transforma-

tions, the ”healthcare industry remains highly suscep-

tible to compromises of valuable health information”

(Chernyshev et al., 2019). Security breaches and data

leakage not only result in reputation and/or ﬁnancial

harm for the healthcare provider/facility but may also

threaten patients’ well-being or health. Weak health

data protection and security measures may result in

detrimental and costly consequences, such as identity

theft, fraudulent insurance claims, ordering drugs for

resale, or even harmful or fatal patient care.

According to a study by Shahriar et al., EHR

applications in general ”suffer from implementation

level vulnerabilities impacting HIPAA requirements,”

and open-source EHR systems are not excluded from

such vulnerabilities (ﬁle manipulation, SQL injection,

possible ﬂow control, etc.) (Shahriar et al., 2021).

However, open-source solutions (OSS) have already

been gaining traction in the scientiﬁc hardware com-

munity for their evidence of cost-effectiveness, and

technological sophistication (Pearce, 2017). In the

healthcare industry, open-source EHR systems are

gaining more attention as adoption rates are increas-

ing, helping to overcome barriers such as excessive

cost and lack of interoperability (Latif et al., 2017).

Besides its cost-effectiveness, open-source EHR sys-

tems offer more ﬂexibility, less vendor lock, and in-

creased control over data; customers utilizing OSS

have more say and control in how data is stored and

used, as compared to proprietary systems. OpenEMR

is a popular and widely used open-sourced EHR sys-

tem and one of the few OSS EHR systems certiﬁed by

the Ofﬁce of the National Coordinator (ONC) of the

US Department of Health and Human Services.

An important application area of the wearable de-

vice is cardiology disease detection. Atrial ﬁbrillation

(Aﬁb) is a quivering or irregular heartbeat (arrhyth-

mia) that can lead to blood clots, stroke, heart failure,

and other heart-related complications. At least 2.7

million Americans are living with Aﬁb. Its prevalence

is 1- 2% of the general population, and it is associated

with increased risk of mortality and morbidity (Behar

et al., 2017). Physicians’ review of the patient’s signs

and symptoms, medical history, and physical exam-

ination including Electrocardiogram (ECG), Holter

monitor, Event Recorder, Echocardiogram, Blood

tests, Stress test, and Chest X-ray are required for Aﬁb

diagonisis. Asymptomatic Aﬁb is more difﬁcult to de-

tect and can go undiagnosed for extended periods of

HEALTHINF 2022 - 15th International Conference on Health Informatics

492

time. Undetected Aﬁb poses more risk to the patient

and may have devastating consequences if diagnosed

too late. Wearable devices are a non-invasive, conve-

nient way to monitor cardiac rhythms, possibly aiding

in the earlier detection of asymptomatic Aﬁb. Aﬁb

detection through wearable devices has the characters

of temporal based trending, adverse event random oc-

currences, and urgent care needed when it happens.

These three characters determine stringent require-

ments of a desired solution: seamless geographic cov-

erage, complex computation support, and low latency,

secure connection. Our proposed 5G edge-centered

telemedicine and decision support system ﬁts the use

case scenario and provides a reliable and scalable so-

lution. Thus, as the proof of concept, we have de-

signed and implemented an Aﬁb detection model de-

ployed to the 5G edge-centered system built and con-

nected to an OpenEMR system. An extension to other

medical models can easily be added to it.

2 SYSTEM DESIGN

2.1 5G Network Framework for

Telemedicine

Figure 1: Hardware Implementation.

Despite the fact that the advantages on physi-

cal coverage of cellular network, IoMT devices in

telemedicine generates a massive amount of data,

some of which may require rapid analysis - such im-

mense and diverse data needs are not supported by the

current 4G/LTE infrastructure (Li, 2019). Latency,

bandwidth, quality of service (QoS), reliability, and

a massive number of connectivity are just some of

the challenges associated with IoT on the current in-

frastructure. Moving forward, telemedicine will need

support for a massive number of devices, standard-

ization, energy-efﬁciency, device density, and secu-

rity (Ahad et al., 2020). To build effective alarm

or decision support models, secure exchange of data

across various platforms is required. The 5G network

is highly attractive in these regards due to its high

speed, massive number connection characteristic, low

latency, ﬂexibility in Radio Area Network (RAN),

and security enhancement due to network slicing for

verticals. Especially, mMTC networks enable the

long battery life,low latency, and high coverage den-

sity with support up to a million devices in a square

kilometer which are crucially for massive scale, ultra-

low-cost hardware.

The implemented hardware platform depicted in

Fig.1 consists of a User Equipment (UE), Base Sta-

tion (BS), Core Network (CN), MEC. The detailed

information for the setup can be referred to at (Wang

et al., 2021).

2.2 MEC Centered Design

Due to wearable devices’ computing and energy con-

straints, telemedicine applications based on IoMT

cannot be executed locally on the terminals, nor

should all health data be uploaded to the cloud for

analysis either. Deploying cloud computing in the

current framework with massive amounts of data gen-

erated by IoT will introduce high data analysis latency

and storage costs, placing tremendous pressure on the

cloud and causing challenges to the network band-

width and end-to-end delay (Zhang et al., 2020). IoT

devices and cloud computing alone cannot fulﬁll the

demands of wearable communication. Instead, mo-

bile edge computing (MEC) should be utilized in tan-

dem with IoT technology, artiﬁcial intelligence, and

cloud computing components.

MEC has characteristics of decentralization, data

localization, and low latency. MEC allows real-time

intelligent decision-making by reducing network de-

lay and transmission costs during the classiﬁcation

and assignment of QoS. By removing most of the

computing needs of the IoT devices, IoT devices be-

come ’lighter,’ essentially only collecting and trans-

mitting data and performing simple computations.

This may save on hardware costs and improve the bat-

tery usage and performance speed of the wearable de-

vice. Utilizing a MEC layer also reduces strain on the

cloud. The MEC Edge layer does not eliminate the

need for a cloud computing layer, but rather the two

layers communicate and work together to enhance the

capabilities of the proposed system. MEC layer han-

dles the pre-processing of data and data analysis be-

fore transmitting data to the cloud for storage and fur-

ther management. The cloud computing layer per-

forms big data analysis, mining, and sharing to train

and upgrade the AI algorithm model that gets pushed

to the edge nodes.

From Wearable Device to OpenEMR: 5G Edge Centered Telemedicine and Decision Support System

493

Figure 2: Proposed Telemedicine System Design.

2.3 OpenEMR-based Distributed

System

A primary concern related to wearable device usage

relates to the security and governance of data sharing.

IoMT generated medical data storing in the central-

ized storage system leads to a single point of failure,

privacy, and security concern (Kumar and Tripathi,

2021). HIPAA protection not extending to wearables

and Apps creates another area of concerns for per-

sonal health data privacy and security. Wearable data

should be stored in secure, regulated personal health

clouds or electronic health records with opt-in sys-

tems, advance security measures, and transparent pri-

vacy policies in place (Bayoumy et al., 2021) in the

desired scenario. In the proposed system design, the

data storage challenges are addressed by a distributed

open source EMR based architecture.

OpenEMR is a widely used open-sourced soft-

ware for electronic health records and medical prac-

tice management solutions, utilized in more than

100 countries worldwide with an estimated usage by

100,000 medical providers serving greater than 90

million patients internationally. OpenEMR leverages

one of the largest communities of users, volunteers,

and contributors dedicated to developing and main-

taining its software, making it a superior alternative

to proprietary counterparts and more comprehensive

than emerging applications. OpenEMR is ONC Cer-

tiﬁed as a Complete EHR, having achieved complete

Meaningful Use certiﬁcation with Release 5.0 and be-

yond. Compared to other popular open-sourced EHR

systems like GNU Health, OpenMRS, and OSHERA

VistA, OpenEMR has the highest functionality and is

among the top for performance (Purkayastha et al.,

2019). Utilizing OpenEMR supports interoperabil-

ity and industry standards and reduces the burden of

seeking new regulation compliance and additional se-

curity measures by making use of a popular exist-

ing electronic health records platform. According to

its website, it offers HIPAA-friendly security features

such as database connection encryption support, ﬁne

grained access control objects, the ability to encrypt

patient documents, and industry-standard password

hashing.

As with other open-sourced software, one of the

most signiﬁcant beneﬁts of OpenEMR in the health-

care industry is that it is free and can easily be down-

loaded from one of the repositories (Syzdykova et al.,

2017). Unlike proprietary electronic medical records

systems, smaller health settings can utilize and adapt

OpenEMR to their needs. Open-sourced systems

are ﬂexible, cost-efﬁcient, offer freedom to try be-

fore buying and avoid vendor lock-in. These ad-

vantages help relieve health disparities by allowing

for greater distribution and accessibility of electronic

health records worldwide.

2.4 Proposed System Architecture

Our proposed system integrates 5G technology with

wearable devices to capture physiological indicators

in real-time and send them to the edge and cloud.

Physiological indicators may include EKG data, heart

rate, oxygen saturation, and other measures captured

via wearable devices. As shown in Figure 2, four

modules constitute the system: IoT module, MEC

module, Network module, and Application module.

Protocols of communications between the modules

are deﬁned. A layer structure is used for the mod-

ule design. Each module is an abstract virtual ma-

chine that provides a cohesive set of services through

a managed interface (Bass et al., 2013). With this de-

sign, layers imbue a system with portability through

the ability to change the underlying computing plat-

form, network, hardware, or application update. The

connection among the four modules is shown in Fig-

ure 3.

Figure 3: System Component Connection.

The IoT Module is responsible for the acquisition

of health-based data and relative ambient data. It is

also responsible for transmitting the data to the in-

HEALTHINF 2022 - 15th International Conference on Health Informatics

494

frastructure, deﬁned by the modules’ protocol. The

category of data acquisition affects the selection of

data transmission. The physical layer data transmis-

sion attribution data is also recorded for security and

acquisition anomaly detection.

The MEC Module is the main component for

deploying OpenEMR instances, anomaly detection,

real-time decision support, long-term data storage,

and data merging to centerized data storage in the

cloud. When a user is an associate with a gNodeB,

the OpenEMR instance will be deployed on MEC

with user customization and relevant user data sync

up from the center OpenEMR location. Acquisition

data from these patients will be transmitted to and an-

alyzed at the MEC in real-time manners. The system

will perform instant decision support, physician assis-

tant alarm, and plan for long-term data backup based

on the results and resources requested. When a user

disassociate with the base station, a clearing up pro-

cess will be launched to release the resources in the

MEC and hand over to the next base station. This

edge based design signiﬁcantly reduce the latency by

bring the data analysis closer to users and avoiding

the routing to core network and enables the real-time

decision support. Firewall and security features lo-

cated on the MEC provide security enhancement to

all instances.

The Network Module includes data transmission

from gNB/MEC to the core network. Trafﬁc security

and efﬁciency is the primary responsibility for net-

work module. The network data are also used for

anomaly detection combined with acquisition data.

Network slicing are used for directing the trafﬁc, iso-

lating context, and enhance network performance.

The Application Module is responsible for cen-

tralized application data management in the cloud.

Long-term analysis and trending, non-real-time ma-

chine learning models are running in the application

module.

2.5 System Implementation and

Automation

To test the system design and implementation, a Ap-

ple Watch Series 6 was selected as the wearable de-

vice during the implementation. ECG, heart rate, and

blood oxygen data were collected. This data from

the Apple Watch is sent via 5G technology to the

MEC layer, where pre-processing and analysis oc-

cur to detect atrial ﬁbrillation. Studies suggest wear-

able devices, like the Apple Watch, may be effec-

tive and convenient tools to diagnose asymptomatic

or symptomatic atrial ﬁbrillation and/or other arrhyth-

mia (Bayoumy et al., 2021). Using artiﬁcial intel-

ligence, MEC determines the classiﬁcation of health

data as either ’trend/normal,’ ’alarm,’ or ’immediate’

and assigns the appropriate QoS resources. If abnor-

malities are detected, and ’immediate’ or ’alarm’ clas-

siﬁcations are triggered, MEC can quickly alert 911

services or medical personnel and the user; this is

the beneﬁt of edge computing - tasks are performed

at the edge of the network, reducing both the dis-

tance of data transmission and communication delay.

Therefore, critical decision-making tasks can occur in

a real-time manner. OpenEMR software was installed

on a Raspberry Pi device and handles the cloud’s stor-

age and management of health data. Raspberry Pi

is a low-cost computer that acts as the server in this

design, continuously running script to handle Apple

Health ﬁles and pre-process then process data for in-

sertion into the OpenEMR database. The MEC layer

works alongside the cloud, pulling relevant data from

existing records in OpenEMR to enhance its artiﬁ-

cial intelligence and pushing data to the OpenEMR

database in the cloud for long-term storage and fur-

ther management. The transmission of data between

the Apple Watch and the MEC layer and to and from

the cloud is performed via 5G technology to ensure

high QoS, low latency, massive connectivity and en-

hanced security. In Figure 4, the ﬂowgraph of the sys-

tem implementation is shown.

Docker

- container system for

OpenEMR, mysql, and

phpAdmin

OpenEMR

process.py:

- imports and reads

ecg.csv and

healthdata.xml files

into dataframes

- formats dataframes

- exports dataframes

as processed.csv

files to local folder

run.sh:

- unzips export.zip

file from shared

folder into ecg.csv

and healthdata.xml

files

- saves ecg.csv and

healthdata.xml

files to local folder

- runs python script

- starts up docker

applications

- removes files from

local folder after

data is uploaded

into database

phpAdmin

mysql

Database

Shared Folder

- contains

export.zip

file

Local Folder

Apple Watch

- paired with cellphone

- data automatically

syncs to cellphone

continuously

RaspberryPi

import.sql:

- reads data from

processed.csv

files

- inserts data into

database

Cellphone

- User chooses when to

export data by

uploading to shared

folder on network

Apple Health

App

- generates

export.zip file

containing

health data

Figure 4: Apple Watch Based IOT connecting to 5G Edge

System Automation.

From Wearable Device to OpenEMR: 5G Edge Centered Telemedicine and Decision Support System

495

3 SYSTEM PROOF OF CONCEPT:

ATRIAL FIBRILLATION

DETECTION IN 5G

TELEMEDICINE NETWORK

In this session, we have designed and implemented

Atrial ﬁbrillation (AF) detection system based on

the previously discussed framework. The model is

trained using data from single-lead ECG plots gener-

ated by AliveCor devices. ECG recordings were col-

lected using the AliveCor device and made available

in (Clifford et al., 2017). A database of 8528 single-

lead ECG and their annotations were used for training

and testing. Four categories of ECG recordings were

present in the databases: atrial ﬁbrillation (A), nor-

mal sinus rhythm (N), other rhythms (O), and noisy

recordings. Using this as proof of concept in the pro-

posed system, we focused on the detection of type A:

atrial ﬁbrillation detection.

This is a proof of concept for the 5G Edge Cen-

tered Decision Support System with OpenEMR For

Wearable Devices. Figure 5 shows the ﬂowgraph of

the Aﬁb detection system itself. Figure 6 offers the

integration and deployment of the system shown in

Figure 5 to the 5G Edge Centered Telemedicine and

Decision Support System. When a user device asso-

ciates to the 5G network, the gNB that the device is

associated with or in the process of handing over to

will start creating the OpenEMR instance with the as-

sociated patient records and the Edge App instance.

Relevant trained non-near-real time models are trans-

ferred from the data network to MEC on Edge App.

As shown in Figure.6, Feature extraction, feature-

based Aﬁb detection model, R Cycle sample-based

Aﬁb detection model will be transmitted to the Edge

app from the data network to MEC. Context data, in-

cluding location, weather, road condition, etc., will

also be accessible by the Edge App. As patient data

are transmitted to the gNB and MEC, the data will be

instantly processed by feature extraction, Aﬁb model

detection, and generating the results. One of the three

potential types of results will be generated, Aﬁb De-

tected, Low Detection Conﬁdence, and normal result.

The Aﬁt detected result needs immediate attention,

with an alarm being sent with the highest QoS. The

low detection conﬁdants need a physician’s decision

and assist with the second level of QoS. The normal

results will be saved for long-term monitoring. As

the patients disassociate with the current gNB(A) or

handover to the next gNB(B), the process will be re-

launched in the next gNB(B); meanwhile, the existing

patient data and Edge App will be removed from the

current gNB(A).

Figure 5: Flowgraph of Aﬁb Detection.

Preprocessing is needed to extract features from

ECG plots. The cardinal features of atrial ﬁbrilla-

tion are an absence of coordinated depolarization of

the atria (absence of P waves on the ECG) and unpre-

dictable depolarization of the ventricles (no pattern to

R wave occurrence on the ECG). As shown in Figure

1, The P wave represents the depolarization of the left

and right atrium and corresponds to atrial contraction,

and the QRS complex includes the Q wave, R wave,

and S wave.

In general, some features are used to describe

ECG medically. The commonly used features are

ECG Signal quality, Heart Beats / Cardiac Cycles,

and Heart Rate Variability (HRV). In HRV, there are

a set of parameters used to describe the ECG sig-

nal, including CVSD, HF, LF, RMSSD, Shannon,

Power, Triang, ULF, VHF, VLF, cvNN, madNN,

mcvNN, meanNN, medianNN, pNN20, pNN50 and

dNN. These are the basic features we used in our

model.

8528 data samples are used for analysis. The

length of each data sample is 9,000 to 18,000, record-

ing 30 seconds to 60 seconds of ECG. Figure 7 shows

an element overlap illustration of when aligning with

the peak of the R wave of a normal ECG. Figure 8

shows an element overlap illustration of when align-

ing with the peak of the R wave with AF detected.

For the data processing part, we ﬁrst categorize

ECG plots by extracting essential Cardio and HRV

features from the ECG. If we can extract them, the

ECG plots are categorized as class 1. If the features

are not extractable due to the low quality or abnormal

ECG, then the plots are categorized as class 2. Here,

we use a python library called NeuroKit (Dominique

Makowski, ) for the feature extraction.

Results show that there are 1.2% class 2 ECG

plots and 98.8% class 1 ECG plots. For class 1 ECG,

we use 27 features extracted from Table 1 to prepare

for the data. We can reach an F1 score of 0.76.

For class 2 ECG plots, we need to take a closer

look at them. Among the 104 ECG plots, there are

HEALTHINF 2022 - 15th International Conference on Health Informatics

496

Figure 6: Flowgraph of Aﬁb Detection Over the Proposed Telemedicine System.

Figure 7: Normal ECG R Wave Overlap Illustration.

Figure 8: ECG with AF R Wave Overlap Illustration.

5% normal ECG, 13% AF, 66% Other Abnormal, and

15% noise. 95% of them are abnormal ECG plots.

This will explain the reason why the features can be

extracted. For the 5% normal ECG, the quality of

the ECG plots also shows low. The experimental re-

sults are shown in Figure 10. The real-time AFib de-

Figure 9: Result of The AFib Detection Model.

tection from the apple watch is validated in our sys-

tem. Depending on the watchOS version, two types of

data sources are supported - a pdf format image and

the raw data ﬁle. As shown in Figure 10, when the

available data source is a pdf image, our system con-

verted it into an accepted data format for the detection

model.

4 CONCLUSIONS

This study proposed a 5G mobile edge computing

(MEC) based telemedicine design integrating wear-

able devices with an Open-EMR electronic health

records system. This design has multiple modules:

the IoT module, MEC module, Network module, and

Application module. A near-real-time artiﬁcial intel-

ligence (AI) components and electronic health record

From Wearable Device to OpenEMR: 5G Edge Centered Telemedicine and Decision Support System

497

Figure 10: Result of Apple Watch ECG image detection.

(EHR) instances are deployed to the MEC layer, en-

abling cloud computing capabilities on the network

edge. 5G technology further improves the latency

and connectivity necessary to support wearables and

IoMT in telemedicine. A proof of concept imple-

mentation of atrial ﬁbrillation (Aﬁb) detection with

frequency predictable by trending, adverse event ran-

dom occurrence, and urgent care needed when hap-

pens are evaluated. Future work includes applications

in telemedicine beyond Aﬁb detection and further de-

velopment of the telemedicine work with mmWave

and integration with other technologies.

ACKNOWLEDGEMENTS

This work was funded by the Commonwealth Cyber

Initiative (CCI), research, innovation, and workforce

initiative of the Commonwealth of Virginia.

REFERENCES

Ahad, A., Tahir, M., Sheikh, M. A., Ahmed, K. I., Mughees,

A., and Numani, A. (2020). Technologies trend to-

wards 5g network for smart health-care using iot: A

review. Sensors (Switzerland), 20(14).

Bass, L., Clements, P., and Kazman, R. (2013). Software

Architecture in Practice Second Edition Third Edition.

Bayoumy, K., Gaber, M., Elshafeey, A., Mhaimeed, O.,

Dineen, E. H., Marvel, F. A., Martin, S. S., Muse,

E. D., Turakhia, M. P., Tarakji, K. G., and Elshazly,

M. B. (2021). Smart wearable devices in cardiovas-

cular care: where we are and how to move forward.

Nature Reviews Cardiology.

Behar, J. A., Rosenberg, A. A., Yaniv, Y., and Oster, J.

(2017). Rhythm and quality classiﬁcation from short

ECGs recorded using a mobile device. In Computing

in Cardiology, volume 44.

Chernyshev, M., Zeadally, S., and Baig, Z. (2019). Health-

care Data Breaches: Implications for Digital Forensic

Readiness. Journal of Medical Systems, 43(1).

Clifford, G. D., Liu, C., Moody, B., Lehman, L. H., Silva,

I., Li, Q., Johnson, A. E., and Mark, R. G. (2017).

AF classiﬁcation from a short single lead ECG record-

ing: The PhysioNet/computing in cardiology chal-

lenge 2017. In Computing in Cardiology, volume 44.

Dominique Makowski. NeuroKit.

Haghi, M., Thurow, K., and Stoll, R. (2017). Wearable

Devices in Medical Internet of Things: Scientiﬁc Re-

search and Commercially Available Devices. Health-

care Informatics Research, 23(1).

Kumar, R. and Tripathi, R. (2021). Towards design and

implementation of security and privacy framework

for Internet of Medical Things (IoMT) by leveraging

blockchain and IPFS technology. The Journal of Su-

percomputing, 77(8).

Latif, S., Qadir, J., Farooq, S., and Imran, M. (2017). How

5G Wireless (and Concomitant Technologies) Will

Revolutionize Healthcare? Future Internet, 9(4).

Li, D. (2019). 5G and intelligence medicine–how the

next generation of wireless technology will recon-

struct healthcare?. Precision Clinical Medicine, 2(4).

Min Wu and Jake Luo (2020).

https://www.himss.org/resources/wearable-

technology-applications-healthcare-literature-review.

Pearce, J. M. (2017). Emerging Business Models for Open

Source Hardware. Journal of Open Hardware, 1(1).

Purkayastha, S., Allam, R., Maity, P., and Gichoya, J. W.

(2019). Comparison of Open-Source Electronic

Health Record Systems Based on Functional and User

Performance Criteria. Healthcare Informatics Re-

search, 25(2).

Shahriar, H., Haddad, H. M., and Farhadi, M. (2021). As-

sessing HIPAA Compliance of Open Source Elec-

tronic Health Record Applications. International

Journal of Information Security and Privacy, 15(2).

Sun, H., Zhang, Z., Hu, R. Q., and Qian, Y. (2018). Wear-

able communications in 5g: Challenges and enabling

technologies. IEEE Vehicular Technology Magazine,

13(3).

Syzdykova, A., Malta, A., Zolfo, M., Diro, E., and Oliveira,

J. L. (2017). Open-Source Electronic Health Record

Systems for Low-Resource Settings: Systematic Re-

view. JMIR Medical Informatics, 5(4).

Wang, Y., Gorski, A., and da Silva, A. (2021). Development

of a Data-Driven Mobile 5G Testbed: Platform for Ex-

perimental Research. In IEEE International Mediter-

ranean Conference on Communications and Network-

ing.

Zhang, Y., Chen, G., Du, H., Yuan, X., Cheriet, M., and

Kadoch, M. (2020). Real-time remote health moni-

toring system driven by 5G MEC-IOT. Electronics

(Switzerland), 9(11).

HEALTHINF 2022 - 15th International Conference on Health Informatics

498