EMPIRICAL SURVEY OF HSPA NETWORKS TO DELIVER

MEDICAL APPLICATIONS

Christopher I. Iyawe

1,2

and John Doucette

1,2

Department of Mechanical Engineering, University of Alberta, Edmonton, T6G 2G8, Alberta, Canada

TRLabs, Edmonton, Alberta, Canada

Keywords: ICT, HSPA, Homecare, eHealth, Healthcare ICT.

Abstract: Population explosion coupled with a decline in medical personnel are forcing governments in the developed

and developing nations to look at new ways to deliver healthcare to their citizens. The cost of providing

healthcare put so much pressure on the delivery system that many find it difficult to access the system. The

most affected are often the elderly and/or those who are highly immobile, particularly those who reside in

remote communities. This group is often laden with chronic illnesses and diseases requiring home care, and

ambulatory services to access the healthcare system. Such services are generally quite expensive to provide.

Developing an eHealth network to drive applications to deliver healthcare services to homes is very costly.

Such dedicated networks will also be difficult to manage. We can observe the manner in which other

industries use information and communications technology (ICT) through the public network to influence

ways services are delivered to the public, and ask why the same (i.e., use of ICT applications through the

public network) cannot be done for the healthcare system. In this paper, we assess the capability of local

HSPA public networks to deliver healthcare applications to home care clients.

1 INTRODUCTION

It has become clear that wireless communications

are almost essential to maintain an active lifestyle in

today’s developed countries. Economic, social and

cultural activities around the world are now being

driven by communications technology. The late

comer to this vast access of communication

networks is the healthcare industry (Fischer, 2010).

Most modern healthcare communications still use

century technology through the public switched

telephone network (PSTN). Although most of the

patient-related information generated in the

healthcare industry is still on paper, advances have

been made in the implementation of electronic

medical records (EMR) in many modern healthcare

clinics and hospitals (Ludwick and Doucette, 2009).

However, there is still much to be desired in the use

of information and communication technology (ICT)

to deliver healthcare services.

Many other industries have successfully used

ICT to deliver services to the home for decades. For

example, the banking industry uses ICT to deliver

account management, bill payment, loans, and other

services (Hardy, 2010). It is only natural to expect

that ICT can do the same for the healthcare industry.

Applying ICT to healthcare is a complex issue,

because healthcare delivery may involve a variety of

applications, each of which has different bandwidth

and latency requirements. For the purpose of this

study we will consider applications that are

applicable to homes, such as vital signs monitoring,

educational health services, etc.

Many healthcare applications require real-time

services, and people’s lives may depend on the

response time of the healthcare provider, which

might in turn depend on the network’s latency and

other technical characteristics for successful delivery

of such applications (Cisco, 2007). This could have

grave repercussions if the network and the

applications that run on it are not designed properly;

bandwidth and latency requirements for such

networks and applications are of utmost importance.

They have to be carefully designed and measured in

order for the healthcare industry to appreciate the

capability and the expediency of such applications.

Where these real-time services are required,

applications with even small latency can have a

disproportionately adverse effect on the user’s

overall experience.

390

I. Iyawe C. and Doucette J..

EMPIRICAL SURVEY OF HSPA NETWORKS TO DELIVER MEDICAL APPLICATIONS.

DOI: 10.5220/0003784203900396

In Proceedings of the International Conference on Health Informatics (HEALTHINF-2012), pages 390-396

ISBN: 978-989-8425-88-1

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

1.1 Goals and Motivation

The purpose of this study is to assess the capacity of

public wireless networks, more specifically local

high speed packet access (HSPA) networks, in

relation to healthcare applications requirements as

they share channels with other types of public

traffic. This work looks at various ICT applications

with potential for use in home care, and determines

their latency and bandwidth characteristics over an

HSPA wireless access network.

We have chosen public wireless network

technology because of its wide reach in suburban

and rural areas where home care is more likely to be

a major aspect of a patient’s healthcare regimen. We

considered HSPA in particular because it is the

fastest mobile wireless network that has been widely

deployed by our regional telecommunication

companies. With fast data access, mobile

communication can provide personal or interactive

services virtually anywhere in the service area at

virtually any time.

With the introduction of HSPA in WCDMA

networks, Internet service providers (ISPs) now

have the capability of offering latency sensitive and

simultaneous voice, data and video services over

their mobile wireless networks. We have collected

empirical data over real networks, analyzed the data,

and have come up with recommendations for

applications that can be run through such networks.

This paper is organized in the following order.

The methodology used in the survey is described in

Section 2. The test setup is given in Section 3.

Survey results are presented in Section 4. Network

evaluation for medical data sets is considered in

Section 5, and the paper finishes with some

concluding remarks in Section 6.

2 METHODOLOGY

We chose Strathcona County, near Edmonton,

Alberta as our test site, as that region covers a good

mix of suburban and rural residential areas. We first

mapped out the test area as illustrated in Figure 1.

The survey area is representative of the population

density distribution in the suburban and rural areas

of Strathcona County. We drove through the

indicated route (highlighted in red) and collected

global positioning system (GPS) data, and wireless

information such as received signal strength

indication (RSSI). We also carried out bandwidth

testing at numerous points in the area to simulate

healthcare applications through the network.

Figure 1: Area of interest for Wireless measurement.

3 TEST SETUP

Figure 2 illustrates our HSPA network setup. We

connect to the radio access network (RAN) with an

access service network antenna to create the data

path between the mobile station and the core

network. Detailed network characteristics and

configurations of the core network are unknown to

us, as the local service providers do not like to

divulge this information. All data presented herein is

based on the data we were able to collect through

our own experimental apparatus.

We had the ability to measure the network

performance from two end points, in the field and in

our office in downtown Edmonton In the field, we

measured the upload speed and latency as well as the

download speed and latency through the network.

We set up a laptop computer (denoted as “Server

PC” in the figure) to act as a fixed server running

Darwin streaming server and IxChariot endpoints

(Ixia, 2010) at our Edmonton office. The Server PC

communicated with our field apparatus (the lower

half of the figure) through the HSPA network. We

deemed the Server PC to represent a doctor’s office

or hospital server and the field apparatus to represent

a patient home. The Server PC has an external IP

address to by-pass the office firewall and also to free

it from our office’s network traffic. This ensures that

our test is isolated from whatever is happening

within the office network.

The “Test PC” is the main computer in the field

apparatus (set up in the lead author’s vehicle), and

ran VLC application software and IxChariot

endpoints. VLC was used in the field to stream

video files from the Server PC. IxChariot was used

to test the upload and download speed through the

network. The Test PC was also fitted with a

Logitech QuickCam and a Polycom PVX. This

enabled us to do video conferencing with our

EMPIRICAL SURVEY OF HSPA NETWORKS TO DELIVER MEDICAL APPLICATIONS

391

office’s video conferencing system via the Server

PC. Video files of various sizes, frame rate,

resolution, and bit rates recorded in MPEG-4 format

were loaded on the Server PC. We deemed the video

files represent medical data images such as

radiographs, MRI, etc. These files were used to test

the network from Test PC in the field.

Figure 2: HSPA Network test setup.

The Test PC was also connected to a Cypress

CTM-152 cellular modem, shown in Figure 2.

Embedded in the CTM-152 is a Sierra Wireless

AirPrime MC8700 HSPA+ module that accepts an

approved data card from local ISPs. The CTM-152

is equipped with antenna (also shown in the figure)

and GPS inputs so that data could be mapped to the

corresponding GPS coordinates at the time of

collection. Low-level diagnostic information about

the network was also collected and logged in

MySQL database.

The measurement route is shown in Figure 1.

Starting at point A in northeast Edmonton, we drove

to the hamlet of Sherwood Park, which is the

suburban area bounded by rectangle B-C-H-J in

Figure 1. Then, we proceeded to point D

(Ardrossan), and ended at point F (Josephburg). At

the noted points and intersections along the route,

we became stationary and ran our battery of network

tests: IxChariot FTP-get and FTP-put, VLC video

streaming and Polycom Video Conference sessions.

These points were chosen to coincide with

community boundaries, community centres, and

places that represent the population distribution in

the area. We followed this route on two different

dates (28 May 2010 and 04 June 2010), and

simultaneously collected data for each of two main

wireless carriers in the region (we will simply refer

to them as Carrier 1 and Carrier 2). We could not

find a convenient place to stop and collect data at

point C as there was construction on the road way at

the time of our measurements.

4 RESULTS/DISCUSSION

4.1 Network Performance – Suburban

We used IxChariot, and VLC to assess the capacity

of the network of the two carriers and we measured

the performance of their networks. We configured

IxChariot to download data files from the server

using FTP-get. These measurements were done on

the networks of both carriers at pre-determined

locations. The first measurement location was at

point B in Figure 1. The RSSI recorded at this

location was -81 dBm for carrier 1. We completed

the first test by downloading a 100 kb data file in

real time. The average download speed was 0.557

Mbps. Figure 3 through Figure 5 give a graphical

view of TCP throughput, transaction rate, and

response time for our FTP-get test on Carrier 1. The

average transaction rate and response time were

0.697 per second and 1.436 seconds, respectively.

As expected, we can observe that a high throughput

tends to correspond to high transaction rates and low

response time. For example, Figure 3 shows a fall in

throughput at about 51 seconds into the test, and

there was a corresponding low transaction rate and

high response time at that time stamp in Figure 4

and Figure 5. The fall in throughput below the

application bandwidth requirement threshold may

necessitate packets drops, just as in the video

streams test. These are important facts to note about

the network when designing an eHealth application

because round trip time of an application will

depend on the network response time, which is the



Antenna

Mode

Server P C

Data Logger

Test PC

HEALTHINF 2012 - International Conference on Health Informatics

392

delay through the network. Also, the delay through

the network will depend on the network throughput.

The ideal for a high capacity network will be a high

throughput, high transaction rate, and low response

time. This will ensure minimal delay between a

doctor’s office and a patient’s home.

Similar values were obtained for FTP-put

throughput, transaction rate, and response time at

point B, where we uploaded a 100 kb data file from

the field to the Server PC. We deem this to represent

a patient sending personal or medical data from

home to a doctor’s office.

Figure 3: FTP-get throughput for Carrier 1 at junction of

Trans Canada 16/Hwy 216 (point B).

Figure 4: FTP-get transaction rate for Carrier 1 at junction

of Trans Canada 16/Hwy 216 (point B).

Figure 5: FTP-get response time for Carrier 1 at junction

of Trans Canada 16/Hwy 216 (point B).

We note here that although Figure 3 through

Figure 5 correspond to only a single point in our

tests, the results shown are typical of data collected

at other locations, and so for space considerations

we omit figures showing the data for those tests.

That said, we show summary data for some location

in Table 1.

The packet size for IxChariot was set at a

constant 4096 bytes, with no dynamic variation in

packet size. Real time streaming protocol (RTSP)

was used in VideoLAN (VLC) streaming of the pre-

recorded MPEG-4 video clips. These files were

created using FFmpeg on a pre-recorded video clip

varying the key parameters as detailed in Table 2.

During the test, the received video streams were

recorded and the video quality evaluated

subjectively by noting audible and visual artefacts in

the audio and video streams that are not present in

the originals.

At point H, we repeated the video streaming tests

and observed that here, we were able to successfully

stream files 1-7 on Carrier 1 without noticeable

degradation, but could only do so with files 1-4 on

Carrier 2. In fact, the video bit rate of Carrier 1 was

5000 kbps, approximately 6 times that of Carrier 2,

which was 750 kbps. We can also note that at this

location, throughput for carrier 2 was much lower

than for Carrier 1. For instance, Carrier 1 averaged

1.084 Mbps throughput for FTP-get while Carrier 2

averages 0.441 Mbps. The average transaction rate

for Carrier 1 was 1.378 per second, and the average

response time was 0.726 seconds, while the average

transaction rate for Carrier 2 was 0.551 per second,

and the average response time was 1.816 seconds.

Given Carrier 2 transaction rates this low, response

times this high and a low throughput rate, it took

almost twice as long to run this test in carrier 2 than

in carrier 1. This suggests that depending upon

choice of carrier (i.e., its network configuration,

number of active users etc.), will greatly impact the

type of eHealth application that can be deployed in

such a network. We show all video streaming

summary data in Table 3, below. In the entire

suburban area, we found out that at a minimum, we

can stream video file 3, a video recorded at 15

frames per second with a bit rate of 500 kilobits, and

screen resolution of 720 X 480. We conclude that at

least based on these findings, we can expect a home

care application including video with those

characteristics to be functional over the two HSPA

wireless networks we tested.

EMPIRICAL SURVEY OF HSPA NETWORKS TO DELIVER MEDICAL APPLICATIONS

393

Table 1: Summary of FTP-get and FTP-put data at all suburban locations.

Location Carrier

RSSI

(dBm)

FTP-get

Throughput

(Mbps)

FTP-get

Transaction

Rate (sec

-1

)

FTP-get

Response

Time (sec)

FTP-put

Throughput

(Mbps)

FTP-put

Transaction

Rate (sec

-1

)

FTP-put

Response

Time (sec)

B 1 -81 0.557 0.697 1.436 0.805 1.009 0.991

G 1 -75 0.956 1.198 0.836 0.764 0.957 1.044

H 1 -76 1.084 1.378 0.726 0.845 1.059 0.944

J 1 -91 0.890 1.114 0.897 0.777 0.973 1.028

B 2 -81 0.456 0.561 1.798 0.351 0.438 2.289

G 2 -91 0.425 0.532 1.891 0.341 0.429 2.304

H 2 -76 0.441 0.551 1.816 0.349 0.346 2.293

J 2 -91 0.417 0.522 1.915 0.348 0.435 2.297

4.2 Network Performance – Rural

As mentioned earlier, our test region includes a large

area with a higher concentration of people in the

suburban (Sherwood Park) area, and a more

moderate population density in the acreages and

rural areas, with the exception of a somewhat higher

density of people in the immediate vicinity of points

D and F in Figure 1 (the rural hamlets of Ardrossan

and Josephburg with populations just over 400 and

200, respectively). With this in mind, we chose test

points in the acreages and rural areas (points E, K,

and L) and in the two above-mentioned hamlets. For

expediency, we will discuss our findings on Carrier

1 here. We could not obtain any useful data for

carrier 2 at point E, K and L. Also, Carrier 1 has

poor signal reception at point E and L. At those

points, we could not run our battery of test. FTP-get

test executed for carrier 2 at Josephburg, but the

signal reception was so poor that FTP-put could not

execute. We summarize the above data and the other

collected data in the rural locations in Tables 4 and

Table 2: Characteristics of MPEG-4 video files used.

File #

Frame Rate

(frames per

sec)

Audio

Bit Rate

(kbps)

Screen

Resolution

Video

Bit Rate

(kbps)

1 10 16 320x240 250

2 10 64 320x240 1000

3 15 64 720x480 500

4 15 64 720x480 750

5 15 64 720x480 1500

6 15 64 720x480 2000

7 15 64 720x480 5000

8 24 128 1920x1080 3000

9 24 128 1920x1080 4000

10 24 128 1920x1080 15000

5 NETWORK EVALUATION FOR

CLINICAL DATA SETS

One of the purposes of conducting this study was to

determine what sort of capability exists in local

HSPA networks (Palola et al., 2004) to

accommodate potential healthcare applications.

Table 6 shows various sizes of data files produces by

some medical applications and the average

transmission time based on our measured averages.

Medical application file sizes are well documented

in various peer-reviewed articles the file sizes used

in Table 6 were adapted from (Rafael et al., 2006).

The frequency of monitoring vital signs, the

number of patients being monitored and the number

of monitoring devices may generate large amounts

of data and traffic for the network. The extent of

monitoring and the frequency of monitoring will

depend on the medical history and medical problem

of the patient (Varshney, 2006). The results of our

measurements show that HSPA has system

resources that are sufficient to handle a shared

transport channel. Vital signs monitoring, which has

a low bandwidth requirement, may easily be

transported through this medium to a server at the

clinician or hospital system. Vital signs devices that

are real-time or near-real-time, running store and

forwarding protocols may work well with this

transport system. Specific tests may still be done to

properly size the number of users on the shared

channel. Bandwidth-hungry applications such as

radiographs can best be served by store and

forwarding systems. As shown in Table 6, it may

take some time to upload or download 100MB file

through the system.

HEALTHINF 2012 - International Conference on Health Informatics

394

Table 3: Summary video streaming results at all suburban locations.

Carrier 1 Carrier 2

Location

Best

File

Best File

Frame Rate

(fr/sec)

Best File

Screen

Resolution

Best File

Video Bit Rate

(kb/sec)

Best

File

Best File

Frame Rate

(fr/sec)

Best File

Screen

Resolution)

Best File

Video Bit Rate

(kb/sec)

B 5 15 720x480 1500 4 15 720x480 750

G 7 15 720x480 5000 - - - -

H 7 15 720x480 5000 4 15 720x480 750

J 7 15 720x480 5000 3 15 720x480 500

Table 4: Summary FTP-get and FTP-put data at all rural locations.

Location Carrier

RSSI

(dBm)

FTP-get

Throughput

(Mbps)

FTP-get

Transaction

Rate (sec

-1

)

FTP-get

Response

Time (sec)

FTP-put

Throughput

(Mbps)

FTP-put

Transaction

Rate (sec

-1

)

FTP-put

Response

Time (sec)

D 1 -90 0.899 1.127 0.888 0.820 1.027 0.973

F 1 -97 0.523 0.657 1.522 0.682 0.854 1.172

K 1 -71 0.932 1.166 0.857 0.723 0.905 1.105

L 1 -102 0.054 0.068 14.722 0.075 0.093 10.742

D 2 -91 0.163 0.204 4.902 0.098 0.123 8.160

F 2 -92 0.059 0.075 13.415 - - -

Table 5: Summary video streaming results at all rural locations.

Carrier 1 Carrier 2

Location

Best

File

Best File

Frame Rate

(fr/sec)

Best File

Screen

Resolution

Best File

Video Bit Rate

(kb/sec)

Best

File

Best File

Frame Rate

(fr/sec)

Best File

Screen

Resolution

Best File

Video Bit Rate

(kb/sec)

D 6 15 720x480 2000 None

F 3 15 720x480 500 None

K 7 15 720x480 5000 None

L none None

Table 6: Medical files upload times in select areas.

File Type File Size

Point G

(Mean TX:

0.845Mbps)

Point D

(Mean TX:

0.820Mbps)

Point F

(Mean TX

0.682Mbps)

Sonogram 2MB 18.9 sec 19.5 sec 23.5 sec

Angiogram 70MB 11.05 min 11.38 min 13.69 min

SPECT 10MB 1.58 min 1.63 min 1.96 min

MRI 30MB 4.73 min 4.88 min 5.87 min

Polycom PVX software was used to test the link

between the video conferencing system at our

downtown office and the Test PC video conference

in the field. Polycom implemented an enhanced

video (H.264) standard. We observed that at the low

bandwidth setting of 384 Kbps, the frame drop and

the video jitter were at a minimum. For example, in

Ardrossan, where RSSI was -90dBm, we recorded a

frame rate of 15 fps, a video jitter of about 7ps, an

audio jitter 14, and a packet loss of 1 for the entire 5

minutes of video calls. This was typical of most sites

on our test route.

6 CONCLUSION

We were able to characterize the RSSI in the study

area and did some channel condition testing in the

predetermined sites. The fact that HSPA is a highly

deployed technology shows the acceptance of the

technology in the wireless community, which would

easily permit a means to obtain high speed data

connection to homecare patients. With this

technology, ISPs have been able to boost their

bandwidth for better delivery of content through

their network and better user experiences with the

wireless technology. We were able to establish a

threshold for video conferencing (384Kps) given the

same channel conditions and same setting on the

Polycom PVX, that good reception and

communication is possible over the HSPA network.

Our study indicates that it is plausible to set up

patient remote monitoring using the public HSPA

wireless network; high quality medical data may be

delivered over the network with the existing network

resources. The low response times we obtained from

our measurement will work with live monitoring

EMPIRICAL SURVEY OF HSPA NETWORKS TO DELIVER MEDICAL APPLICATIONS

395

applications such as ECG, SpO

. With live

monitoring application, the most important service

requirement will be the response time or latency

which affects the quality of the eHealth application

over the network. The network will perform better

with a store and forward application, where medical

information is stored at the time of collection and

then later forwarded to the server at a convenient

time (e.g., when there is sufficient network

resources to do so). This provides value to the

patients and the clinicians in the sense that the

clinician can still monitor the patient and have a

sense of the patient’s wellbeing. Using the system

relieves and frees up clinical resources that would

otherwise have been tied up in the process of

gathering these medical information.

REFERENCES

Cisco, 2007. Design Best Practices for Latency

Optimization, white paper. Cisco Systems, Inc.

Fischer, R., 2010. Influencing of ICT on national and

global development, available online (accessed on 08

November 2010): www.connect-world.com/

index.php/component/k2/item/3224-influence-of-ict-

on-national-and-global-development.

Hardy, I., 2010. KPMG: 15% of Canadians have done

banking through a mobile device, available online

(accessed on 18 August 2010): mobilesyrup.com/

2010/08/18/kpmg-15-of-canadians-have-done-bank

ing-through-a-mobile-device/, 2010.

Ixia, 2010. IxChariot, available online (accessed August

18, 2010): www.ixchariot.com/, 2010

Ludwick, D. A., Doucette, J., 2009. Primary Care

Physicians’ Experience with Electronic Medical

Records: Barriers to Implementation in Fee-for-

Service Environment, International Journal of

Telemedicine and Applications. Hindawi.

Palola, M., Jurvansuu, M., Korva, J., 2004. Breaking

Down the Mobile Service Response Time,

International Conference on Networks (ICON 2004).

IEEE.

Varshney, U., 2006. Enhancing Wireless Patient

Monitoring by Integrating Stored and Live Patient

Information. Computer-Based Medical Systems

(CBMS 2006). IEEE.

Maia, R. S., von Wangenheim, A., Nobre, L. F., 2006. A

Statewide Telemedicine Network for Public Health in

Brazil. Computer-Based Medical Systems (CBMS

2006). IEEE

HEALTHINF 2012 - International Conference on Health Informatics

396