MIT IEEE URTC Proposal for HealthVisor

HealthVisor: A Look into Data-rich Bio-monitoring
Brian Goldwyn1
, Andrew Pastore1
, James Spencer1
, Weihui Li2
, Chen-Hsiang Yu1
1
Computer Science and Networking 2
Biomedical/Medical Engineering
Wentworth Institute of Technology Wentworth Institute of Technology
{goldwynb, pastorea1, spencerj1, yuj6}@wit.edu liw1@wit.edu
Abstract - The rapid development of technology has made real-
time telemedicine possible, yet about 735,000 Americans have
heart attacks annually and 43% of sudden cardiac deaths
occurred outside hospitals. Survivors often suffer from anxiety
and depression, which makes the prognosis worse. In this paper,
we propose a new system, HealthVisor, to address this issue. The
goal of this research is to develop a real-time, low-cost health
monitoring system to provide the patients’ electrocardiography
(ECG) and stress level data to medical professionals. HealthVisor
consists of four parts: (1) Careband Wearable Device (CWD) - a
custom wearable device to collect heart data from the patients;
(2) Careband Android Application (CAA) - an Android application
to receive data from CWD and pass it to the back-end Web
service; (3) HealthVisor Web Server (HVS) - a back-end Web
service to process, analyze and store the patient’s health data,
and; (4) HealthVisor Front-end (HVF) - a Web interface to
present analyzed data from HVS. The prototype system was
evaluated on three subjects, and the results indicated that the
collected data are reliable and the system is promising for further
clinical study.
Keywords - heart disease; wearable system; monitoring
I.INTRODUCTION
Heart diseases have become one of the leading causes of
mortality around the world [4]. According to the Centers for
Disease Control and Prevention (CDC), there are about
735,000 Americans having heart attack annually. In a 2005
survey, most respondents (92%) recognized chest pain and
about 47% of sudden cardiac deaths occur outside the
hospitals [4]. Therefore, early action and awareness is critical
for the detection of a heart attack. In a 2008 study, Home
Monitoring technology allowed early detection of atrial
fibrillation in patients with pacemakers or implantable
cardioverter defibrillators (ICDs) [10]. In 2006, approximately
280,000 pacemakers and 160,000 ICDs were implanted in
America [13]. Holter monitor has been and still is a standard
system to monitor cardiovascular disease remotely in past
decades [14]. The disadvantage of the commonly available
Holter monitor include: (1) the patients may not have
symptoms during the period of recording; (2) physicians can
review the data only after the patient returns the Holter
monitor. Lately, a real-time wireless Holter monitor has been
developed and proved by US FDA [3]. More recently, many
wearable real-time ECG monitor systems were developed [2,
6, 7] that can monitor for an extended duration.
Stress, though not immediate fatal, is another factor that
affects people of all ages and genders. In 2011, 22 percent of
Americans report extreme stress: an 8-10 on a 10-point scale
where 1 is little or no stress and 10 is great deal of stress [1].
Galvanic skin response (GSR) has been traditionally used to
monitor stress levels [11]. Picard’s research group at MIT has
developed a wearable device to monitor stress [8].
When patients with cardiovascular disease experience anxiety
or depression, the situation becomes worse [9]. To the best of
our knowledge, there is no wearable device that can
simultaneously monitor the patients’ ECG and stress level at
the same time. Simultaneous monitoring ECG and stress level
has a significant impact to cardiovascular disease management
which can shift the role of ECG monitor as a pure diagnostic
device to diagnostic & therapeutic device, as the increase of
stress level will alert the patient to adjust their emotion
positively to avoid serious cardiovascular diseases.
The objective of this study is to develop a new system to help
medical professionals access patients' ECG and stress data
while either party are not in the hospital. We propose to have a
new system, HealthVisor, to solve this problem. HealthVisor
contains four different components: a wearable device, a
mobile application, a back-end Web server and a front-end
Web interface. The details of each component are introduced
in the following section. In addition, technical challenges,
conclusion and future work are also discussed respectively.
II.COMPONENTS
HealthVisor relies on four major components. The first one,
Careband Wearable Device (CWD), is a health monitoring
device that uses sensors to gather physiological data from the
user. This data is passed to the second component, Careband
Android Application (CAA), a smartphone application which
processes and relays the data to the third component,
HealthVisor Web Server (HVS), which is a cloud server. This
server stores the data in a database and waits for HealthVisor
Front-end (HVF), the fourth component, to retrieve it. (see
Figure 1).
Figure 1. Simple Data Flow
A. Careband Wearable Device (CWD)
CWD was developed on an Arduino microcontroller
system for prototyping purposes. It contains multiple sensors

to collect health data, including Electrocardiography (ECG),
Galvanic Skin Response (GSR) sensor and Temperature
sensor. While ECG sensor provides an in-depth reading of the
electrical signals from the heart, GSR sensor measures the
electrical conductance of the skin as an indicator of
psychological or physiological arousal. We specifically use
GSR sensor to measure the user’s stress level. In addition, the
CWD uses a Temperature sensor to monitor the patient's body
temperature. To support the communication between CWD the
CAA, an Adafruit nRF8001 Bluetooth Low Energy (BTLE)
breakout was used to transmit sensor data (Figure 2). The
wiring and module configuration of the Arduino are
illustrating in Figure 3.
Figure 2. Careband Wearable Configuration
Figure 3. Careband Wearable Device (CWD) prototype
B. Careband Android Application (CAA)
CAA is a relay service. It receives collected data from
CWD via Bluetooth and reformats this data to make it ready
for upload to the HealthVisor Web Server (HVS). Since
security is a concern in medical domain, CAA is HIPAA1
-
compliant. Therefore, CAA component can securely upload the
data from the wearable device to the server using HTTPS.
Upon launching the app, users must login – the credentials are
securely passed and verified by the HVS. After successful
login, the app will show the user their ECG data as it comes
in. The main purpose of this functionality is not for data-
1
The Health Insurance Portability and Accountability Act (HIPAA)
is a law signed in 1996 that sets regulations for many topics in the
healthcare industry, including how software should handle
information to aid digital security.
analysis, but instead to provide graphical feedback to the user.
At the beginning of the design, the app was simply targeted at
providing data relay functionality between CWD and HVS.
However, along with the development, the scope of the
application has been extended to visualize low-accuracy ECG
data (Figure 4) and execute as a background service for
running multiple applications simultaneously on the phone.
Figure 4. (Left) Careband Android Application (CAA); (Right)
A Real-Time ECG Data Visualization
C. HealthVisor Web Server (HVS)
The HVS component is designed to manage user accounts
and authentication as well as provide a scalable repository for
health information sent by the CAA. Basically, the CAA sends
the patient’s data securely via HTTPS to the HVS and the data
is decrypted and stored in the database. The HVS handles user
login and registration from both the phone (CAA) and
HealthVisor Front-end (HVF). The server does not store
passwords in plain text, but instead generating a SHA512 hash
for the password and stores the result to improve the security.
While the HVS is hosted by Heroku and operates through
NodeJS, we use MySQL as a database hosted by JawsDB. The
HVS not only handles all website requests, but it also performs
encryption and decryption for HTTPS TLS. This is also where
data is processed and analyzed to extract useful knowledge
from the data.
D. HealthVisor Front-end (HVF)
The front-end component, HVF, provides an integrated
interface for medical practitioner to easily read the health data
of a large number of patients. Specifically, it provides options
to monitor the patient's health data graphically. An important
component of the HVF is HighCharts [5], a JavaScript module
that allows the Web server to display intuitive and interactive
graphs. Drilldowns, a feature of HighCharts, allows the user to
interact with the charts to expand certain points to get a more
detailed look. This allows a lot of patient data to be shown at a
quick glance while still providing robust depth of data. Figure
5 shows a screenshot of the front-end interface after user
login.

Figure 5. HealthVisor Front-end (HVF)
III. TECHNICAL CHALLENGES
There were many technical challenges encountered during the
implementation phase of the system. In this section, we would
like to summarize them with our thoughts and experiences.
A. Bluetooth Transmission Stopped Sensor Reading
A critical problem encountered while sending ECG data
through Bluetooth (BLE) was that a large portion of ECG data
were missing between each BLE write command. After
investigating the issue in Arduino, it was found that there is a
535ms delay during the execution of a BLE write command,
which writes 5 characteristics of 20-bytes each. Because
Arduino runs every command synchronously, the main loop
designated for reading sensor information is also delayed
during this interval. It turns out that many vital ECG readings
were missing while the Arduino processed BLE
communication.
To solve this problem, sensor reading was moved out of the
main loop and into an Interrupt routine (via TimerOne external
library). This meant that there is now an interrupt every 1ms
that triggers the reading of sensors, even during when the
Arduino is being delayed to send BLE information.
B. Data Loss from BTLE Transmission
When analyzing the ECG data sent from CWD to CAA, we
found that entire 20-byte chunks of data were missing
frequently. The initial thought was that CAA fails to process
the Bluetooth characteristics correctly. Upon inspection of the
Arduino serial monitor, it was seen that errors occurred
whenever a characteristic failed to be sent from CWD to CAA.
The problem was that Arduino sent 20-byte characteristics too
frequently, and the phone running the CAA was sometimes
incapable of handling so many transmissions within such a
short time interval. To solve the problem, a modification was
made to Adafruit_BLE_UART.cpp file to change the delay
time from 35 milliseconds to 50 milliseconds between each
characteristic write. This effectively decreased data loss
through BLE transmission between two devices.
C. Asynchronous ECG Graphic
A complicated problem occurred when we tried to
asynchronously update the ECG graph on the CAA without
destroying the data to be uploaded to the remote server (HVS).
A simple list iteration of the data would not suffice because
the list is constantly growing from asynchronous threads that
collect the Bluetooth data and shrinking from posting data via
HTTP to the server (HVS). Also, the webview, which displays
the graph via Smoothie [12], could only be run in the main UI
thread. This added an extra complexity since CAA manages
the UI using Fragments, which change the logic of how to
access the UI thread. In addition, since CAA implements
Bluetooth Low Energy (BLE) as a background service, it is
not guaranteed that the UI is visible.
This was solved by creating a custom method in the
Fragment, which can be called and passed data from the
background service. The custom method accesses the UI
thread of the Fragment and passes data to the webview's
graphing engine. It was noted that the graphing engine
(Smoothie) fails to display data at the rate that CAA reads it
from BLE, causing an inaccurate graph.
D. Memory Overflow
The most challenging issue in the CAA is memory overflow. It
is caused by the Java Garbage Collector that terminates critical
variables used by the CAA, including the lists of sensor data.
This happens when the application uses ~16 megabytes in
local variable storage. However, the collection of sensor data,
including ECG, GSR and Temperature, accounts for ~7
megabytes or less per hour. The garbage collection can occur
with inconsistent timing – anywhere from one second after
app launches to 10 minutes or more. Additionally, the local
variables are being emptied in every 12 seconds when the CAA
posts data to the server. All threads and the background
service are checked if they already exist before being created.
Further analysis and investigation are needed to solve this
problem.
IV. CONCLUSION AND FUTURE WORK
Clinical data indicates that an easy-to-use heart and stress
monitoring system is important for the patients and their
medical practitioners [9]. In this paper, we develop a portable
heart and stress monitoring system, HealthVisor. The system
provides a real-time ECG and stress level monitoring
functionality for the patients and their medical professionals
alike. The system was targeted to be used outside the hospital.
The collected data from CWD can be accessed in the
HealthVisor Front-end (HVF) from anywhere with an Internet
connection. This is an important step for incorporating
mobility into medical care.
Although HealthVisor system has demonstrated a promising
direction for mobile medical care, there are still ways to

improve and extend the system. Following is a summary of
future work for this research.
A. Offline Access
The healthcare industry is the most common target for
cybercrime. It's not uncommon for a hospital's network to go
down, causing doctors to lose visibility on their patient's
history. An offline version of HealthVisor is considered as a
desktop application. It will securely locally store all data
relevant to the medical practitioner’s patients. It would
constantly update its database from the HealthVisor Web
Server (HVS). All charts and historical health data can still be
accessed, even during a network failure.
B. Extendibility to Compatible Devices
To expand HealthVisor, it is necessary to make it
compatible with pre-existing equipment, such as Fitbit or
Jawbone. Additionally, to decrease the size and increase
portability of CareBand, it is useful to explore a greater
variety of micro-controllers other than Arduino Uno, such as
Lilypad.
C. Real-time Health Alerts
It would be useful for medical practictioners if the CAA
can recognize abnormal patterns in the data and alert the user
when potential heart related issue detected.
ACKNOWLEDGEMENT
This research project was partly sponsored by Wentworth
Presidential EPIC Mini-Grant, which was awarded to an EPIC
course. EPIC learning is a new experimental learning program
at Wentworth Institute of Technology. It is an externally
collaborative, project-based, interdisciplinary curriculum for
learning. We also appreciate Prof. Leonidas Deligiannidis for
encouraging the research and development of this project.
REFERENCES
[1] Anderson, N. B., PhD, Johnson, S. B., PhD, Belar, C. D., PhD,
Breckler, S. J., PhD, Nordal, K. C., PhD, Ballard, D., MBA, . . .
Dukes, S. (2012, January 11). Stress in America: Our Health at
Risk. American Psychological Association.
doi:10.1037/e506172012-001
[2] Burri, H., & Senouf, D. (2009). Remote monitoring and follow-
up of pacemakers and implantable cardioverter defibrillators.
Europace, 11(6), 701–709.
http://doi.org/10.1093/europace/eup110
[3] Dolan, B. (2010, December 15). FDA clears Curvus wireless
ECG monitor. Retrieved September 02, 2016, from
http://mobihealthnews.com/9757/fda-clears-curvus-wireless-
ecg-monitor
[4] Heart Disease Facts. (2015, August 10). Retrieved September
02, 2016, from http://www.cdc.gov/heartdisease/facts.htm
[5] Interactive JavaScript charts for your webpage | Highcharts.
(n.d.). Retrieved September 02, 2016, from
http://www.highcharts.com/
[6] Linder, M., & Kaib, T. (2004, January 20). Patent US6681003 -
Data collection and system management for patient-worn
medical devices. Retrieved September 02, 2016, from
https://www.google.com/patents/US6681003
[7] Lobodzinski, S. S., & Laks, M. M. (2012). New devices for very
long-term ECG monitoring. Cardiology Journal Cardiol. J,
19(2), 210-214. doi:10.5603/cj.2012.0039
[8] Sano, A., Picard, R. W., "Stress recognition using wearable
sensors and mobile phones", The 5th biannual Humaine
Association Conference on Affective Computing and Intelligent
Interaction (ACII 2013), 2-5 September 2013
[9] Frasure-Smith, N., & Prince, R. (1985). The ischemic heart
disease life stress monitoring program: Impact on mortality.
Psychosomatic Medicine., 47(5), 431-445.
doi:10.1097/00006842-198509000-00003
[10] Ricci, R. P., L. Morichelli, and M. Santini. "Remote Control of
Implanted Devices through Home MonitoringTM Technology
Improves Detection and Clinical Management of Atrial
Fibrillation." Europace 11.1 (2008): 54-61. Web.
[11]Shi, Y., Ruiz, N., Taib, R., Choi, E., & Chen, F. (2007).
Galvanic skin response (GSR) as an index of cognitive load.
CHI '07 Extended Abstracts on Human Factors in Computing
Systems - CHI '07. doi:10.1145/1240866.1241057
[12] Walnes, J., & Noakes, D. (n.d.). Smoothie Charts: A JavaScript
Charting Library for Streaming Data. Retrieved September 02,
2016, from http://smoothiecharts.org/
[13] Wilkoff, B. L et al. HRS/EHRA Expert Consensus on the
Monitoring of Cardiovascular Implantable Electronic Devices
(CIEDs): Description of Techniques, Indications, Personnel,
Frequency and Ethical Considerations: Developed in partnership
with the Heart Rhythm Society (HRS) and the European Heart
Rhythm Association (EHRA); and in collaboration with the
American College of Cardiology (ACC), the American Heart
Association (AHA), the European Society of Cardiology (ESC),
the Heart Failure Association of ESC (HFA), and the Heart
Failure Society of America (HFSA). Endorsed by the Heart
Rhythm Society, the European Heart Rhythm Association (a
registered branch of the ESC), the American College of
Cardiology, the American Heart Association, 2008. Europace,
10(6), 707-725. doi:10.1093/europace/eun122
[14] Wimmer, N. J., Scirica, B. M., & Stone, P. H. (2013). The
Clinical Significance of Continuous ECG (Ambulatory ECG or
Holter) Monitoring of the ST-Segment to Evaluate Ischemia: A
Review. Progress in Cardiovascular Diseases, 56(2), 195-202.
doi:10.1016/j.pcad.2013.07.001

MIT IEEE URTC Proposal for HealthVisor

More Related Content

What's hot (20)

Similar to MIT IEEE URTC Proposal for HealthVisor (20)

Recently uploaded (20)

MIT IEEE URTC Proposal for HealthVisor