Quality Evaluation of Ultrasound Medical ImagesTransmission over UTRAN

WENBING YAO1, ROBERT. S. H. ISTEPANIAN1, ABDUL. Z. SALEM1 ALVARO ALESANCO2,HARRIS ZISIMOPOULOS3, PHIL GOSSET3

1Mobile Information Engineering and E-Med System Group, Department of Electronics and ComputerEngineering, Brunel University, London, UB8 3PH, UK

2 Department of Electronics and Communications Engineering, University of Zaragoza, Spain3Group Research and Development, Vodafone Group, Newbury, RG14 1JX, UK

AbstractA 3G wireless solution for the OTELO system -- a tele-echography robotic system -- is

introduced in this paper. Based on the analysis and understanding of the OTELO functional modalities,the IP Multimedia Subsystem (IMS) is introduced in the 3G wireless solution to the OTELO system.Quality of Service (QoS) issues for tele-robotic control and tele-echography image transmission arediscussed and evaluated. We investigate the throughput performance of the system with a channelsimulator that can investigate and demonstrate the effects of a UMTS radio link on multimediaapplications in terms of user perceived quality of service (QoS). Two types of traffic are transmitted withdifferent source coding and QoS schemes over the UTRAN (UMTS) system simulation. The results showthat the best throughput is reached when transmitting the less highly compressed image data. With Eb/N0of -0.6dB, the real average throughput can reach 333.5 Kbit/sec whilst the theoretical channel bandwidthis 384 Kbit/sec.

Key Words: UTRAN, Multimedia traffic, Tele-echography, Tele-robotic, Mobile medical service, 3G mobilecommunications system, mobile telemedicine

1. INTRODUCTION

It is well know that ultrasound echography is a popular easy-use non-invasive method for routineexaminations in hospital in different diagnostic settings. Ultrasound imaging in medicine can be used toevaluate the degree of emergency for patients. However the main drawback of current ultrasoundtechniques is that the quality of the examination depends highly on the operator's specialised skills, skillsthat are lacking in most health centres, especially in many small medical centres, isolated sites, or rescuevehicles. In some cases, no appropriate well-trained sonographer can be found to perform the firstultrasound echography from which the emergency level can be evaluated[1].

Because of this concern, a fully integrated end-to-end mobile tele-echography system that willbring population groups preventive care support is studied and developed in the OTELO project. OTELOis a dedicated, portable ultrasound probe holder robotic system, associated with new communicationtechnologies, which will reproduce from afar the expert's hand movements during an ultrasoundexamination. Although being manipulated by a non-specialist on the remote site, the slave system willbring, in real time, good image quality back to the expert site, where force feedback control will becombined with virtual reality for the rendering of the distant environment. Recent works have shown thatit is possible to transmit realtime ultrasound images over ISDN and IP networks with a bandwidth of 384Kbis/sec, thus permitting ultrasound specialist to be located at other centres different from the centrewhere the images were acquired[2]. Although wired networks could offer a reasonable solution for theinterconnection of the OTELO† system in some scenarios, many times like in rural areas or in rescuevehicles, there is no available wired network that could offer the bandwidth required by an ultrasoundsystem.

† This work is supported by the EU supported project: Mobile Tele-Echography Using an Ultra-LightRobot (OTELO). Project Contract no.: IST-2001-32516.

The concept of ‘Unwired E-Med’ that will enable any person to communicate informationregarding their health and medical needs with complete mobility, anytime and anywhere is anamalgamation of the unwired world, which will flourish within the first decade of the 21st century. Thelaunch of the Third Generation Mobile Systems (3G) will have a huge impact in this process, because itwill directly address the problems associated with most critical ‘mobile telemedicine’ systems and healthcare services that are not readily supported by existing mobile technologies. As the Third generation (3G)mobile radio networks like Universal Mobile telecommunications System (UMTS) is now beinglaunched, and will support bit rates of up to 144 Kbit/sec in rural areas, 384 Kbit/sec in hotspots and up to2 Mbit/sec in indoor scenarios that is much higher than presently available in second-generation cellularsystems, UMTS is considered as a wireless communication solution for OTELO system[1] yielding to amore general solution. In comparison with satellite or ISDN, 3G wireless communication technology canprovide a cheaper and more convenient mobile medical service for the OTELO system[3].

Mobile medical services, which provide medical information, and help healthcare organizationsto deliver mobile medical services to the public is an important field, and one with a potential market ofseveral billion £/ year. It is also an emerging field which will have a revolutionary impact on the deliveryof medical care worldwide. Consumers will benefit from the high quality of health services available, andthe affordable costs of the mobile medical services.

In Europe, wideband CDMA (WCDMA) was chosen as the air interface of UMTS by theEuropean Telecommunications Standards Institute (ETSI). During the past decade, extensive researchwork has been done on WCDMA systems. Most of these studies focused on the issues of transmittinggeneral multimedia traffic over 3G channels. A comprehensive overview of the various mobiletelemedicine technologies as used in different settings is given in [3]. However, in medical environments,different data traffic with different classes of QoS requirements have to be transmitted simultaneously.Considering the importance of the medical services within the mobile communications market, it is veryimportant to study the 3G network performance for this particular application. One of the critical issuesthat have to be investigated is how to optimize the network performance to provide medical services withother multimedia services simultaneously.

In this paper, the OTELO system is presented and the transmission of ultrasound images fromthe OTELO system to an expertise centre is evaluated. To perform the evaluation, a UMTS simulatordeveloped by Vodafone has been used. This simulator allows us to modify the characteristics of themobile channel and the transmission rate so as to evaluate different scenarios in terms of throughput andglobal time delay in the transmission of ultrasound images using FTP protocol.

The paper is organized as follows: in section 2 the user and medical service functionalrequirements for OTELO wireless communication system are described. The OTELO UMTS wirelessstructure identified for OTELO functional modality is illustrated and QoS issues for OTELO medicalservice as well as the Vodafone UMTS simulator are discussed and presented respectively in section 3.Simulation results of the system throughput performance are provided and discussed in section 4 beforewe conclude in section 5.

2. ULTRASOUND AND OTELO SYSTEM

The operation of OTELO system is subdivided into the following functional modalities:

2.1 Medical diagnostic Procedure

In the OTELO system, a medical consultation session starts by the medical expert at the mastersite communicating in video and voice format with the remote patient at the isolated slave site. The expertmay also adopt wired (ISDN, ADSL or LAN) or wireless access (UTRAN) means if available. Thewireless access is the solution adopted here. During this phase, the correct positioning of the probe byvoice comments to the isolated operator is first calibrated, and the probe will be allocated in parallel to therobot main axis (perpendicular to the skin) that is used as the reference position to start moving the robot.As soon as the probe is well located the “video conference image” can be interrupted and the isolated

slave station sends the first echo images (real time high resolution sequences). The following sequentialscenario’s can then be used:

1) If as expected the organ to be visualised appears on the echo image, the expert switches to highresolution echo images and uses the distant robot to adjust the probe orientation on the patient and to findthe appropriate incidence needed for the diagnosis.

2) If the desired organ does not appear on the image, the expert asks the operator to translate the probeholder system (“left, right, up, down”) without patient-probe video control until the expected organappears on the echo view. Then he continues the sequence as described above.

3) If the image of the organ cannot be reached by the previous translations, the expert will asks theassistant at the slave station to reposition the probe via the video conference system (back to thebeginning of the sequence).

All views (real time dynamic sequences) will be stored at the master station (namely, expertcentre) first, and then the expert can determine if the images need to be stored on the image data server,which is located somewhere on the internet or intranet, for either on-line or off-line co-operative workwith other medical experts. Figure 1 shows the OTELO system architecture and operation mode, asexplained above.

INTERNETINTRANET

Slave stattion located at thesecondary hospital or isolated area

Video-Conference

Ultrasound device

Probe holder robotic system

Robotcontrols

Forcesensors

UMTS

Master station locatedat the expert center

Medical expert

Video-Conference

Ultrasound images

Forcefeedback

Robotcontrols

Datastorage/Imageserver

Figure 1 OTELO system.

2.2 Robot Real Time Controls

OTELO is a medical robot that posses a 6-degree-of-freedom (DOF). Since the change inorientation of the robot probe must be done in a continuous mode, and the velocity of these movementsmust be adjustable from the operating centre, the robot feedback control data is very sensitive to timedelay. Current research shows that a time delay of less than 20ms is acceptable for the teleoperation of arobot over the Internet[5]. It is important to promise a constant data rate for robot feedback control, if therobot control data is sharing the bandwidth with other types of data, such as videoconferencing images,etc. about 50% of the connection bandwidth should be assigned to robot control to keep the control loopactive. In comparison with TCP, connectionless protocols such as UDP, RTP (Real Time Protocol) or E-UDP (Enhanced UDP) are suggested for adoption to provide quicker data delivery. The operationrequirements in terms of bandwidth needed for teleoperating the robot are shown in Table I.

Control command Operationalfrequency of robot Estimated data rate

6+6 floats for jointposition/velocities frommaster to slave plus a longinteger for time stamping

50-100kbit/s frommaster to slave

6 floats for force sensorreadings from slave to masterplus time stamping

100-200Hz 20-50kbit/s fromslave to master

Table I Requirements for real time control data for 6DOF robot.

2.3 Ultrasound Images

Echographic images coded with DICOM (Digital Imaging & Communications in Medicine)standard as well as JPEG-LS, JPEG-2000 as well as ultrasound diagnostic video will be used in OTELO.The size of the medical images varies from 256x256 pixels up to 512x512 pixels while the grey level ofeach pixel can be saved up to 16 bits. The lowest data rate acceptable to medical experts is 210 Kbit/sec.During the beginning of the examination, when the expert is searching for a specific examination area(liver, kidney...), a high quality ultrasound image sequence is not required. An image with 20 grey levelscould be sufficient at this stage to ensure that the received images can be considered to be of acceptablequality by the expert. During the second phase of examination, when the organ of interest has been found(e.g. heart observation), sequences of high-resolution ultrasound images will be needed. Due to thequantity of information that has to be transmitted is high, compression techniques could offer a significantreduction in the transmission rate in the case of ultrasound diagnosis video or in the time needed tocomplete a transmission in the case of still echographic images[10, 11, 12]. The compression ratesachieved by these methods can reach factors of 20:1. Unfortunately, sometimes it is necessary to considerreversible compression techniques (e.g. Lempel–Ziv algorithm[6, 7] or JPEG[8, 9]) that would bring highquality images to the expert yielding to a minor compression ratio, usually ratios of 3:1. As the ultrasoundimages and ultrasound diagnosis video are mostly transferred from the robot probe or image data server tothe master, an asymmetric channel is applicable. Table II shows the requirements for the transmission ofultrasound diagnosis video and still ultrasound images needs are shown in Table III.

Lossy Lossless

Bit Rate (bit/s)

Kbps: 64K,mean{PSNR}: 37.48 dBKbps: 128K,mean{PSNR}: 37.98 dBKbps: 256K,mean{PSNR}: 38.77 dB

>2M

Frame Rate (frame/s) 30 N/AColor / B&W Grayscale N/A

UltrasoundDiagnostic

Video

Resolution 320x240 N/ATable II Requirements ultrasound diagnosis video.

Image Type Image Size(pixel2)

CompressionRatio (bpp)

Uncompressed 8Lossless 4.96Near lossless(85dB) 4.2

ImagesWith Text

Lossy(35dB)

768x576

0.77Uncompressed 8Lossless 2.99Near lossless(85dB) 2.73

ImagesWithoutText

Lossy(35dB)

512x512

0.44Table III Requirements for still ultrasound images.

2.4 Ambient Information

This information mainly deals with the ambient video and audio information and patient’s reflexand video feedback to the OTELO video operation during the medical diagnosis procedure. The timedelay between the slave probe motion (remote controlled probe) and the reception of the video imagefrom the slave station at the expert centre should be as close as possible to 300 ms. The suggested framerate for video transmission is 15 frame/s to minimum 1 frame/s. Colour or B/W video will be chosen intransmission. Interactive compression standards, H.263 or MPEG-4 will be used to transmit the real-timevideo. (MPEG-4 coding involves nearly 5sec delay although providing good quality image). Table IVshows the requirements needed for the transmission of ambient information.

Frame Rate (frame/s) 25~30 N/ASize (pixel2) QCIF: 176x144 N/AAmbient

Video-conferencing Compression

Standards MPEG4 or H.323

Table IV Requirements for ambient video-conferencing.

2.5 UMTS traffic classes selection

We have defined the bandwidth requirement of the OTELO system in their differentfunctionality modes. From these requirements, the classification of the OTELO traffic is mapped to thethree major traffic classes defined by the 3GPP UMTS QoS Classes[13]. In summary, the ambient videoand audio can be treated as stream class traffic and ultrasound image as interactive class traffic, whilstrobotic control data must be treated as conversational class traffic since it’s very sensitive to time delay.These traffic classes will be presented in the next section.

3. UTRAN STRUCTURE AND VODAFONE UMTS SIMULATOR

3.1 System architecture

In this section, a UMTS mobile communication solution is proposed to provide the OTELOmedical service with appropriate QoS. The IP Multimedia Subsystem (IMS), which has recently beendeveloped in UMTS release R5[4] as an extension to the existing Circuit Switched (CS) and PacketSwitched (PS) domains, is introduced in the solution to support the OTELO medical service. Theconfiguration of the system entities, which consists of mobile terminal, UTRAN radio access network,GPRS evolved core network, the specific functional elements of the IP Multimedia Core Network (IMCN) subsystem, and OTELO application server (OTAS), is illustrated in figure 2. The medical serviceoffered by the OTELO system is a value-added service (VAS) for the mobile network operator. TheOTELO Application Server (OTAS) is the main element to control this medical service via S-CSCF,especially for the QoS control while transferring the OTELO data package. The Internet is aconglomeration of networks utilizing a common set of protocols. IP protocols are defined in the relevantIETF STD specifications and RFCs. The networks’ topologies may be based on LANs (e.g. Ethernet),Point to Point leased lines, PSTN, ISDN, X.25 or WANs using switched technology (e.g. SMDS,ATM)[5]. The OTELO Medical Data Server (OMDS) in the Internet Network is a medical database thatstores the ultrasound medical images, patient profiles and other relevant medical data.CSCF is in charge of session handling and mobility management. It delegates the service control onbehalf of the mobile end-user through a single IP Multimedia Service Control (ISC) interface to OTASand is comprised of three main areas [4]:

� The S_CSCF (Serving CSCF) is the main point for session control within the home network. Itholds user data (downloaded from the HSS), terminates the call/session signaling from themobile terminal (Ue) and interacts with the applications and services area via the ISC interface.

� The I_CSCF (Interrogating CSCF) interrogates the HSS during mobile terminatedcommunications set-up, to determine the S_CSCF catering for the mobile

� The P_CSCF (Proxy CSCF) is closely linked to the GGSN and performs bridging of thesignalling between the terminal (Ue) and the S_CSCF, the proxy is also involved with bearerconfirmation, decision, policy and charging aspects during the session.

The network’s central database Home Subscriber Server (HSS) is an evolution of HLR with theinclusion of additional user data to cater for the IM aspects , and provides subscription and mobility datato the service control entity CSCF and also supplies information for the SGSN. The IM Media Gateway(IM -MGW) acts in similar way to the circuit switched area, providing the transcoders andinterconnection with legacy networks PSTN. As with the OTELO system, the medical expert may alsocommunicate via ISDN, IM-MGW is necessary in this situation. The MGCF is applied in the same wayas for wired VoIP networks in providing IP specific call/session signaling to legacy network signalingconversion (e.g. SIP to ISUP).

Figure 2. System architecture of the UMTS wireless communication solution for OTELO system (bold lines: interfacessupporting user traffic; dashed lines: interfaces supporting only signaling; twin lines: reference point)

Here, the OTELO Master Station could link from either the IP Multimedia Network or from ISDN/ADSLor UTRAN. Figure 1 only illustrates the situation where it links from the IP Multimedia Network. As theultrasound images are mostly transferred from the robot probe to the OTELO Master Station, the airInterface Uu between the OTELO Slave Station and the RNC bears an asymmetric traffic load. Theultrasound image, ambient video & audio steam and robotic control feedback data are transferred overthe uplink channel, while only the latter three types of data are transmitted over the downlink channel.

3.2 QoS issues for OTELO medical service

The UMTS R5 architecture is a layered architecture with a clean split between bearer (e.g.SGSN, GGSN), session (e.g. P-CSCF, S-CSCF) and service level. UMTS QoS is based on hierarchicalbearer layers, as can be seen in figure 3.

Figure 3. QoS Architecture of UMTS[6]

The UMTS QoS bearer service consists of several layered bearer services within the UMTSnetwork. Packet Data Protocol (PDP) Context procedures of the Session Management (SM) protocol willactivate, modify and deactivate the use of UMTS bearers. 3GPP has defined four UMTS QoS classes,also referred to as traffic classes. Table 1 illustrates the different classes with their QoS requirements.According to the definition, the robotic control data falls into the category of Conversational class trafficbecause it is very sensitive to time delay and can only tolerate very small time delay. Ambient video andaudio are normal stream class traffic and Ultrasound image is interactive class traffic. RSVP protocol isused to signal the requested IP QoS end-to-end from a client to a router network and further to a remotehost. The request is per flow and therefore it is claimed to be poorly scalable for backbone networks. Thereservation is done separately for upstream and downstream directions with Path and Resv messages.

Table V: UMTS QoS classes[13]

The Path message is sent from the sender and it includes the traffic characteristics. The Resv message issent from the receiver and it makes the actual resource reservations along the path. As we mentionedabove, the OTELO Master Station may either link from the IP Multimedia Network or through theISDN/ADSL or through UTRAN. The situation in which a laptop at the master side connects to the SlaveStation using RVSP is a typical example of end-to-end QoS scenario 4 that 3GPP specified. (Scenario 4:IP Bearer Service Manager in the User Equipment, end-to-end QoS with IP layer signaling, GatewayGPRS Support Node is RSVP aware .) RSVP here is used to signal end-to-end the required QoS and issent over the radio interface[7]. For the end-to-end scenario in which both the OTELO master station andslave station access through UTRAN, no end-to-end RSVP solution is available in 3GPP, at least so far.This is primarily because RSVP would introduce an overhead on the radio interface, and raises scalabilityproblems when used in backbones both in terms of bandwidth and processing power. In this scenario, to

provide quality control for OTELO medical service there is a need to involve the session layer (SIP/SDP)and co-ordinate bearer and session layer. The way to setup the session is based on the principles ofinterleaving session and bearer level signaling. Since this scenario is not foreseen to occur frequently, weleave the discussion to later papers.

3.3 Vodafone UMTS simulator

In order to evaluate the QoS in the transmission of an image from the OTELO slave station(where the ultra sound images are acquired) to the OTELO master station (where the ultra sound imagesare processed) a UMTS proxy developed by Vodafone has been used. The UMTS Proxy is a channelsimulator that can be used to investigate and demonstrate the effects of a UMTS radio link on multimediaapplications in terms of user-perceived quality of service (QoS). For instance, it can be used toinvestigate interactions with higher layer protocols such as standard IP applications (Web-browsing, FTP,Telnet, etc.). The case we are interested in is the evaluation of a FTP transaction from the slave station tothe master station, as commented earlier.

The UMTS Proxy system comprises two PC workstations, one for the UMTS Proxy itself andone for the UMTS Proxy Configuration & Monitoring Tool (CMT). The UMTS Proxy sits between theclient and server on a LAN segment. Instead of traffic from the client being sent to the server, it isdirected to the UMTS Proxy (by setting the IP destination address). On receiving an IP packet from theclient, the IP Kernel Module within the UMTS Proxy passes the packet down to the UMTS Module forprocessing (IP Address translation and UMTS Radio Link Emulation). The IP Kernel Module checks theIP header of each IP packet arriving at the UMTS Proxy. When the source IP address is either the addressof the client or the address of the server, the packet is re-routed to a special raw socket where it is thenprocessed by the UMTS module. This process is illustrated in figure 4.

UMTS Module

Application Server

Multimedia Data,Control & Signaling

TCP UDPIP

Application Client

Multimedia Data,Control & Signaling

TCP UDPIP

IP Kernel Module

Multimedia Data,Control &

UMTS Proxy

Configuration & MonitoringTool

IP Dispatcher

IP Address Translation

UMTS Radio Link Emulation

LAN LAN

TCP/IP TCP/IP

Figure 4 UMTS Proxy functional schematic

The IP translation function simply translates the source and destination IP addresses such that thepacket can be onward distributed to the server (with the Proxy appearing as the sender client). In thereverse direction the Proxy would swap the destination (Proxy) address for that of the client and changethe source (server) address to that of the Proxy. In both directions, the IP header checksum is alsorecalculated to reflect the new IP addresses to avoid the packet being discarded at the receiving host.

After IP address translation, the packet is passed to the UMTS radio link emulation functionwhich consists of a UTRAN layer 2 and 3 protocol stack implementation with segmentation, schedulingand ARQ and a physical layer bit error injection. A multipath Rayleigh fading channel is assumed. TableVI lists the RLC and PHY parameter settings of the services emulated by the UMTS Proxy. Each serviceis defined by traffic class and maximum data rate. At this moment, the characteristics of the channel(traffic class, maximum data rate, etc.) are selected. Only dedicated channels are considered. This setup ofthe channel will give the error pattern and thus the number of packets containing errors that will bediscarded yielding to a specific QoS. This QoS can be evaluated analyzing the results of the transaction

recorded in the CMT. In this case, the CMT acts like a network analyzer recording all the packetsexchanged between the server and the proxy (both directions) and the client and the proxy (bothdirections).

Traffic Class Conversational Streaming Interactive or Background

Target FER (for TBs) 10-3 10-3 10-3

RLC mode TM TM AMRLC payload size [bits] 640 576 320 320 640RLC header size [bits] 0 0 16TB size [bits] 640 576 320 336 656CRC size [bits] 16 16 16Coding scheme TC TC TCTTI [ms] 20 40 40 20 20 20 10 10Max data rate [kbps] 32 64 28.8 64 128 384 8 32 64 128 384 2048Number of TF per TFS 2 2 3 5 6 8 2 3 4 5 6 7TFS:#TBs per TBS

0, 1 0,2 0,1,2 0,1,2,4, 8

0,1,2,4,8, 16

0,1,2,4,8,16,32,48

0,1 0,1,2 0,1,2, 4 0,1,2,4,8

0,1,2,4,8,12

0,1,2,4,8,16,32

Table VI RLC AND PHY parameter settings of the services emulated by UMTS proxy (TBD: settings for MAC layermultiplexing+shared channels (scheduling)).

4. SIMULATION RESULTS AND DISCUSSION

In order to validate the performances of the OTELO functional modalities represented in Section2, we have carried out the following simulations:

4.1 Simulation of the transmission of still ultrasound images

In this case, still ultrasound images are acquired by a special ultrasound scanner probe, and arecompressed using JPEG2000 with different compression ratios[14]. According to the functionalmodalities of the medical service, listed in Table I, the compression ratio is between 20:1 and a Losslessratio. In the experiments, 34 compressed ultrasound images, of which the original image size is 768x576pixels2 with total file size of 3.86M bytes, are transmitted via FTP as background class traffic in thesimulation environment.

As no errors can be detected at the receiving entity in the transmission process with the TCP/IPstack, instead the Throughput and the RTT (Round Trip Time) are used to evaluate the transmissionperformance. The average throughput of a WCDMA dedicate channel, using different image compressionratios is given in Figure 5 and Figure 6. Two scenarios with theoretical throughputs of 384kbit/s and128kbit/s, whilst Eb/N0 increases from –0.6dB to 0.3dB are considered. Note that in the above figures,compression ratio 0 represents lossless compression.

Figure 5. Average throughhput of images files with different compression ratios, and with different levels of Eb/No, over aWCDMA dedicate channel. The theoretical throughput is 384kbit/s. (Eb/No= -0.6dB; Eb/No= -0.5dB; Eb/No= -0.4dB).

Both Figure 5 and Figure 6 show that the throughput increases as the Eb/N0 rises, but the compressionratio decreases. The first result is obvious, as the higher the value of Eb/N0 is, the better the signal qualityis, and the faster speed the system.

Figure 6. Average throughput of images files with different compression ratios, and with different levels of Eb/No, over a WCDMAdedicate channel. The theoretical throughput is 128kbit/s. ( Eb/No= 0.1dB; Eb/No= 0.2dB; Eb/No= 0.3dB ).

The latter phenomenon can be explained by the TCP congestion algorithm’s influence on throughputefficiency. By changing the length of the congestion window, the TCP protocol tries to adapt the channel.It takes a certain time for the algorithm to track the optimum window length. Since the file sizes of ourcompressed ultrasound images reduce as the compression ratio increases, and the original sizes of theultrasound images are fixed, at 768x576 pixels2 or 512x512 pixels2, so before the congestion algorithmadjusts the congestion window to its optimum length, and the throughput reaches its peak, the filetransmission has already completed 0. Figure 7 compares the total time for transmitting the differentcompressed images with different Eb/No levels. The total transmission time decreased, as the total trafficamount decreases when the images are more compressed.

Figure 7. Total time of transmitting image files with different compression ratio

in different Eb/No level (Eb/No= -0.6dB; Eb/No= 0.1dB ).

4.2 Simulation of the transmission of combined still ultrasound images and full motionultrasound video stream.

The still US images are 16:1 compressed as described in section 4.1. The full motion video istransmitted via Microsoft Netmeeting (H.323). In this case, the US images and videoconferencing aremultiplexed and simulated as stream traffic on a channel with 384kbit/s theoretical bandwidth. The Eb/N0level is –0.6dB. Experimental results show that, when the video window size is “middle size” and theaverage frame rate is 16frames/s, then the average throughput of the multiplexed traffic approaches250kbit/s.

5. CONCLUSIONS

Due to the increasing popularity of ultrasound with GPs and the setting up of a peripateticradiographer-based community services by many hospitals, the movement of using modern informationtechnology to provide ultrasound services via wireless link, both in real-time and stationary mode,between primary care clinicians and specialists at some distances from each other is assured.

This paper proposed an UMTS wireless solution for OTELO system, an end-to-end tele-echography system that integrated a dedicated portable ultrasound probe holder robotic system with newwireless communication technologies. Based on the analysis and understanding of the user and medicalservice functional requirements, the IMS is introduced in the solution to support the OTELO medicalservice. The OTELO UMTS wireless structure identified for OTELO functional modality is illustrated,and end-to-end QoS control for tele-robotic control and tele-echography image transmission arediscussed. The throughput performance of the system over the UMTS radio link was evaluated. Theresults show that the best throughput is reached when transmitting the less highly compressed image data.With Eb/N0 of -0.6dB, the real average throughput can reach 333.5kbit/s whilst the theoretical channelbandwidth is 384kb/s. A balance point between the compression ratio and throughput performance, toreach the optimal transmission performance of OTELO system should be obtained in the future.

6. REFERENCES

[1] ISTEPANIAN (R.), LAXMINARYAN (S.), “UNWIRED, The next generation of Wireless and Internetable TelemedicineSystems-Editorial Paper”, IEEE Transaction on Information Technology in Biomedicine, Vol.4, 3, pp.189-194, Sept. 2000[2] CHAN (F .Y.), TAYLOR (A.), SOONG (B.), MARTIN (B.), CLARK (J.), TIMOTHY (P.), LEE-TANNOCK (A.), BEGG (L.),CINCOTTA (R.) and WOOTTON (R.), “Randomized comparison of the quality of realtime fetal ultrasound images transmitted byISDN and by IP videoconferencing”, Journal of Telemedicine and Telecare, Vol. 8, Nº 2, pp. 91-96, Feb. 2002[3] PATTICHIS (C. S.), KYRIACOU (E.), VOSKARIDES (S.), and ISTEPANIAN (R.S.H)., “Wireless Telemedicine Systems: AnOverview,”, IEEE Antennas and Propagation, Vol.44, 2, pp.143-153, 2002.[4] HOLMA (H.), TOSKALA (A.), “WCDMA for UMTS,” John Wiley & Sons, Ltd, New York, 2000.

[5] OBOE (R.), “Web-Interfaced, Force-Reflecting Teleoperation Systems”, IEEE Transaction on Industrial Electronics, Vol. 48,No.6, Dec. 2001[6] LEMPEL (A.) and ZIV (J.), “Compression of two-dimensional data,” IEEE Trans. Inform. Theory, vol. IT-32, pp. 2–8, 1986.[7] ZIV (J.) and LEMPEL (A.), “A universal algorithm for sequential data compression,” IEEE Trans. Inform. Theory, vol. IT-23,pp. 337–343, 1977.[8] WALLACE (G. K.), “The JPEG still picture compression standard,” Commun. ACM, vol. 34, pp. 30–40, 1991.[9] NIJIM (Y. W.), STEARNS (S. D.), and MIKHAEL (W. B.), “Differentiation applied to lossless compression of medicalimages,” IEEE Trans. Med. Imaging, vol. 15, pp. 555–559, 1996.[10] ERICKSON (B. J.), MANDUCA (A.), PALISSON (P.), PERSONS (K. R.), EARNEST (F.), SAVCENKO (V.), andHANGIANDREOU (N. J.), “Wavelet compression of medical images,” Radiology, vol. 206, pp. 599–607, 1998.[11] RICKE (J.), MAASS (P.), HÈANNINEN (E. L.), LIEBIG (T.), AMTHAUER (H.), STROSZCZYNSKI (C.), SCHAUER (W.),BOSKAMP (T.), and WOLF (M.), “Wavelet versus JPEG (Joint Photographic Expert Group) and fractal compression. Impact onthe detection of low-contrast details in computed radiographs,” Investigative Radiology, vol. 33, pp. 456–463, 1998.[12] MITCHELL (J. L.), PENNEBAKER (W. B.), FOGG (C. E.), and LEGALL (D. J.), MPEG[13] 3GPP TS 23.107, “QoS concept and architecture”, V4.2.0[14] http://www.bourges.univ-orleans.fr/otelo/site[15] TAFERNER (M.), BONEK (E.), “Dwireless Internet Access over GSM and UMTS,” Springer publisher, Berlin, 2001.