Design and Implementation of a Mobile Medical Information System Ying-Wen Bai Department of Electronic Engineering Fu Jen Catholic University Taiwan, R.O.C. e-mail: bai@ee.fju.edu.tw Yung-Sung Huang Department of Electronic Engineering Fu Jen Catholic University Taiwan, R.O.C. e-mail: a9250615@st2.fju.edu.tw ABSTRACT In this paper, we design and implement a mobile medical information that integrates the remote medical videoconference, the real-time electronic medical records and the wireless communication network, which provides users with accessing the through wireless channel and hand-held devices such as, Tablet PC and PDA. Our design use the videoconference compress technology using the MPEG-4 standard, a large amount of shortens video stream data at 1/24 sizes. Each picture or frame is compressed from 153.6 K bytes to 6.21 K bytes, with effective error detection, which balances the quality of the medical pictures and transmission efficiency of the wireless communication network. In addition, our design provides integration of the medical records with images, text, and waveform of the medical information obtained by the database mechanism that is easily stored, searched, transmitting and shared with the authorized medical persons. KEY WORDS Electronic Medical, Video Stream, Error Correction 1. Introduction Recently,videoconferencing used the ISDN (Integrated Services Digital Network), which uses H.320 of CCITT to compress the information. The general users can accept H.320 image quality, but cannot accept the sound quality. Hence, some software methods compress the sound by an individual specification. In order to preserve the quality of H.320 and compatibility, every manufacturer may have different H.320 selections for receiving and decoding. But the selections can cause compatibility problems, such as control of cine-camera, form sharing, and conveying materials. Basically, H.320 only can provide the image and the sound transmission compatibility, but representation compatibility may not be preserved. Unless we use the same selections, otherwise the compatibility may be a problem [1], [2], [3]. Moreover, previous designs only provided text and still pictures of the medical record. Our design provides the dynamic images that are needed in many medical applications such as, ultrasonic images, when the doctor needs to look around over patient's bodies. Otherwise, the doctor can only see a certain picture that is shot at a specific angle. The single picture is not able to see that ultrasonic pictures appearing on the original medical information. This helps the doctor lacking omnidirectional reference material while judging the medical condition. Our prototype provides the video information functions, which use the compression technology by MPEG-4, within a limited bandwidth to obtain a high elasticity and efficiency to exchange the patient's image through the wireless communication network and hand-held devices. We also enhance the electronic medical records, so the patient's instant quantity the physiological data can be stored in one s medical record in the database, enabling one to read them from a distant place through the wireless networks. There are communication channels, client platform, database and servers that are developed by the platform as shown in Fig. 1 [9], [10]. The users at client side use PDA with embedded operating, Tablet PC with Windows, and other hand devices. The communication channels of the can be from outdoor wireless WAN and in hospitals through wire LAN. IBM PC Patient Patient Medical development platform Notebook Order Order PDA Physiological Physiological Database Access Component SQL Database Mobile Phone Streaming Server Datebase Server Mobile Server DICOM Image Information System HL7 Medical Information System Message Legacy System Examine Examine Interface Application Componet Server Fig. 1 Development platform for medical information Customer Platform Develop Platform Protocol Platform 2. The Technological Discussion for Video Systems Our video needs a streaming server and proxy server both using MPEG-4 in respect to the video
broadcast equipments. The video stream server is located the far-end to offer the stream of video service, and responsible to transmit the video stream file to the proxy server at the near-end. The client side uses 802.11b or cell-phone, such as PHS to obtain MPEG-4 compress stream by wireless transmission technology. After MPEG-4 decompressing and decoding at the client site, the user can watch the video stream. In our prototype, the video stream server uses a general PC to perform the video stream to a proxy server, as well as provide MPEG-4 Encoder & Metadata Generator, MPEG-4 Bit stream Transponder, MPEG-4 Key Frame Generator to offer better functions. In the prototype, the proxy server also uses a general PC to offer wireless stream services. 2.1 The Key Modules of the Video Stream Server (a) (b) (c) (d) When there are clients on line, the can set up two carry out threads to each client, which includes the control channel thread and the data channel thread. The server can tabulate and manage the clients to access database through control channel. When the client requests a video streaming, the server starts the service and controls the hybrid ARQ mechanism module, and turn on the packetizer and bitstream modules. In order to meet the user selection for the transmission mode, the server carries out the data channel thread, check each data of the package in order to convey with transmission mode. 2.1.1 Hybrid ARQ Mechanism Because the wireless communication environment has noise interference, the video stream may have more restriction. We must adopt UDP protocol in order to maintain the transmitting efficiency even in unreliable operations which are obtained by the mechanism of correcting errors during the transmitting and processing. In addition, by using the hybrid ARQ mechanism, we adopt Reed-Solomon code, which has the ability to read the n firsthand information pieces of packets, after which the FEC yard will produce m (m> n) convey and send out pieces of package via server. After the client receives n pieces of the packet, it can reduce firsthand information, composite method of ARQ (Hybrid ARQ Mechanism). The server end sends n packet data first, begin and convey parity packets, once recipient accept get n pieces of packets, the server will send out an acknowledge, and the server begin and convey next materials section at once promptly (n pieces of packets). Like this, even if the server does not receive acknowledge, it can continue the processing to accomplish ARQ. 2.1.2 MPEG-4 Bitstream Transcoder Module Fig. 2 shows the transformation of the encoder for the prime field series connection and shape of the frame composition. In each block I represents a reversed direction, Q represents quantilization, DCT represents discrete cosine transform, F represents frame memory, MC represents motion compensation, MV represents motion vector quantity and DS represents down sampling. Input MPEG 2/4 MPEG-4 Decoder IQ1 MV IDCT1 DS F MC MC DCT MPEG-4 Encoder Fig. 2 Transformation of the encoder F Q2 IQ2 IDCT2 Output MPEG 4 simple profile Decode and encode can use different standard (such as MPEG-2 and MPEG-4), such as with different bit-rate, frame-rate, sport vector and MB (Macro block) code attitude. To increase the processing speed of the picture elements, we adopt sport vector and MB code that provide the adjusting ability of the processing speed of the code. By using the picture element to simplify the picture characteristic, the shape controls of the encoding method are described as follows: (1) Reduce the speed of processing picture Because the MPEG-2 standard has a structure of I- B-P (GOP) and picture speed 30 frames per second (fps), but MPEG-4 use B-frame forms, the picture (or frame) speed is 10Hz; therefore the needs to reduce the picture speed. By using each frame head information, we learn whether the decode frame will be B-frame. The will jump over downward frame information, until finding an I-frame to continue decoding. There are two advantages. First, B-frame does not refer to other pictures, and does not affect other frame and degrade the picture quality. Second, in order to increase processing efficiency, the MPEG-2 B- frame movement vector is the bidirectional forecast and compensates the operation of P-frame two times, if there is a jump over the next frame, we can reduce large amount of the computation complexity. (2) Spatial downscaling By using the spatial downscaling, we can reduce the video stream by 1/4 or 1/16. For YUV, within the neighboring we take 4 or 16 values and average in Y, U, and V. In the movement vector aspect, we use a vector median filter. For some N movement
vectors V={V 1,V 2,.,V N }, the mathematical formula is shown as Eq. (1). Vmed V N N Vmed Vl Vi Vl, i = 1,2,..., N...(1) L= 1 K= 1 After Vmed as shown in Eq. (1) is obtained, this value is used to replace the synthesis in the MB movement vector. 2.2 The Key Modules of the Client End The client end establishes the connection for each user's running the, which has three main functions: (1) Control Channel Thread by use of the TCP protocol and the Streaming Video Server. The channel provides a segment channel, picks up the program table which the server provides, as well as the request broadcast movement and feedback to the server hypothesis as well as the network transmission condition. (2) Data Channel Thread and Packet Assembler provide the data channel used for receiving the video stream, the offer the client the order of the video stream, with a multiple broadcast and transmit bit by bit only. After receiving the video stream, the must use the packet assembler and make up a package frame, and then the frame can be decompressed for broadcasting. (3) The modules of the Control Playing, Decoding, Error Detection and Concealment provides broadcast smoothly and avoids the problems of the buffer underflow. Client can monitor the buffer capacity when it receives the data; it takes the control of MPEG-4 decoding. Because the error rate of the wireless network is higher than that of the wired network, we use application layers to control and avoid the packet loss in the Video Coding (VCL) state. Because the package must be checked via CRC and Check sum in the transmission course, when the Application receives the packages that have been corrected. Some of packages might have losses, but the Application can reduce the packet error in wireless network by the error correction mechanism. In the control part of picture playing, we will check every frame is full, or judge whether this frame can be decoded successfully. In addition, the decoder can detect the error, and then fix the packet error or loss. 2.3 Error Detection The different compression standards before MPEG-4 are relied mainly on code structure that contains the Video Objects. There is a lot of Video Object under Video Object. There are 4 small blocks under MB. The mainstream decodes the Video Object Plane. It appears the Video packet utilize VLC to judge the VP block to be solved by using the MPEG-4 specification. Errors in the blocks are detected by calculating the average of the block, and then find out the first block error. 3. Data Flow Diagram Fig. 3 shows the database of the medical information [8]. Four kinds of professionals, such as doctor, nursing staff, patient and information manager will access the database. The doctors use the to access medical records for reading and updating. The nurses need to treat patients according to the medical record daily form for required prescriptions. The manager will maintain the movements of the medical records based on every user's authority. The manager will set up information sub, which includes medical record sub, medicines sub and nurse's database sub. DT doctor 3.0 Medical record Medicial history Diagnosis record Login 2.0 NSS nurse database medical history record Physiological record PT patient remedy 0 Medical information Diagnosis medicine command NS nurse symptom 4.0 MDS medicine database Select medicines Jurisdiction setup 1.0 PIS person input SA analysis Basic document input Fig. 3 The flow diagram of the medical database 4. Error Correction Simulation 4.1 MPEG 4 video transmission simulation under the wireless communication The error detection simulation is realized by using Matlab based on the wireless communication environment establishing, MPEG-4 parameter establishing and I picture error detection and simulation results. 4.1.1 The Wireless Communication Environment Establishing Our prototype uses FEC (Forward Error Correction) in the wireless environment for the (Signal Noise Ratio) = 0~16dB. The error rate of at 16dB may be too high, therefore unsuitable to transfer MPEG-4 video-stream. In order to simulate the feasible channel, we select several kinds of to do the simulation, and the results are shown in Fig. 4 and Table 1.
Table 1 Bit error rate with respect to different (db) FEC(BER) (db) FEC(BER) 0 4.07E-01 16 7.72E-03 8 1.35E-01 18 2.07E-04 12 4.32E-02 20 4.10E-04 BER FEC 1.00E+00 1.00E-01 1.00E-02 1.00E-03 1.00E-04 0 2 4 6 8 10 12 14 16 18 20 (db) Fig. 4 Bit error rate with respect to different 4.1.2 MPEG4 Parameter Establishing The simulation provides four segments of the video-information (Claire, Salesman, Grandma, Container), picture form QCIF in MPEG-4 with 30-frame and different video-information of each one in simulation of I picture. 4.1.3 I Picture Error Detection and Simulation Results 4.1.3.1 Simulation Techniques The video stream is formatted by MPEG-4 code, transmitted the code via 802.11. The receiving end after decoding through 802.11, passing the MPEG-4 decoding, and improves the efficiency of the video stream. 4.1.3.2 Simulation Results and Discussion MPEG-4 format are transmitted in the wireless environment, the error of the channel can be spread over space and time. The index of the error detection has defined the examining data rate (DR), and the frame data rate (FDR). The DR shows the error detection accuracy of examining. The FDR is used for detecting the errors of the frame data. Because I picture including a large amount of picture information behind for P picture, the will imitate an I picture error and can be hidden by inserting the latency inside. The quantity under each picture and different examines the results as shown in Table 2 and 3, or Fig. 5 and 6, include two kinds of No FEC of P (1) on average without the error and hidden the latent picture, (2) totally detection and hide latently. Table 2 The average error detection accuracy for No FEC protection of I picture (16dB) (18dB) DR(%) Pictures Clairs Salesman Grandma Container Average DR 87.95% 84.39% 84.17% 88.07% Average FDR 11.01% 13.23% 9.89% 8.01% Average DR 89.32% 86.09% 89.45% 89.71% Average FDR 7.71% 8.98% 5.05% 6.06% 90.00% 89.00% 88.00% 87.00% 86.00% 85.00% 84.00% No 2/3FEC 16.00 18.00 (db) Claire Salesman Grandma Container Fig. 5 FEC protection picture EC under access (a)claire(b) Salesman (c)grandma(d)container Table 3 The average error detection accuracy for FEC protection of I picture Clairs Pictures Salesman Grandma Container Average DR 91.22% 86.34% 92.11% 92.67% (16dB) Average FDR 6.66% 7.10% 5.09% 4.01% (18dB) DR(%) 96.00% 94.00% 92.00% 90.00% 88.00% 86.00% Average DR 94.89% 91.99% 95.82% 92.98% Average FDR 1.96% 1.98% 1.88% 3.01% 2/3FEC 16.00 18.00 (db) Claire Salesman Grandma Container Fig. 6 I picture has 2/3FEC under the protection the EC simulation by (a)claire,(b)salesman,(c)grandma, and(d)container 4.1.4 Wireless Access to Packet Loss Network For the case of packet loss network, we choose the packet size L b = 47 bytes, the same as ATM payload with one byte for sequence. If we use the Internet packet size, the proposed video streaming can be adopted for Internet video stream. The average packet loss rate (P 1 ) of testing packet is from 0 to 10%, and the average burst length is
fixed to 5 packets. We use the Reed-Solomon codes with channel block size =128 bytes and the allowed redundancy for error protection is fixed to 10% of the total bandwidth. Fig. 7 shows the average P results and average throughput over 30 simulations under different packet loss rates. The simulation shows the performance with pre-interleaving is much better than that of the without pre-interleaving. For the case of P 1 =0.1, the with pre-interleaving achieves more than 2dB gain in P and about 40% improvement in throughput comparing with the without pre-leaving. When the average packet loss rate P 1 decreases, the difference between the with and without pre-interleaving becomes smaller. This is because more lost packet can be recovered via channel coding in a low packet loss rate. Therefore, the proposed preinterleaving is more efficient when the network experiences moderate and high loss rate. we should fine- tune the storing devices, set up the videoinformation in PDA with the quality of 320X240, transferring 6 frames per second. The X-Ray image for the electronic medical record Video Conference System Administration Fig. 8 Medical information appear at a Tablet PC P(dB) 29 27 25 23 21 19 17 15 0.00 0.01 0.02 0.05 0.08 0.10 Packet Loss Rate with preinterleaving without preinterleaving Supersonic waveforms of the dynamic images (a) Throughput (%) 100 90 80 70 60 50 40 0 0.01 0.02 0.05 0.08 0.1 Packet Loss Rate with preinterleaving without preinterleaving (b) Fig. 7 The comparison of pre-interleaving under different packet loss rates (a) P (b) Throughput 5. Experimental Results The prototype includes two sets of client ends. The first one as shown in Fig. 8 is developed on Tablet PC; the instant picture of video-stream can be launched larger to 640X480, transfer 24 pictures or frames per second for electronic medical record, including X-Ray images and supersonic waveform files. The second one is developed on PDA as shown in Fig. 9 that the user accesses the ultrasonic file of a patient's belly. Because the operation ability of PDA is different from PC, so we will broadcast the pictures ahead. To broadcast smoothly, Fig. 9 Medical information appear at a PDA 6. Conclusion In this paper, we design and implement the mobile medical information by the integration of the software and hardware from the wireless communication, PDA, Tablet PC, and the wire networks. Due to the consideration of the power consumption, we choose the PHS as the wireless communication. The power consumption of the PHS is lower than that of the GSM. In addition, the bandwidth of the PHS is larger than that of GSM by 6 times. Overall, the PHS may not interfere with the medical apparatus signals; therefore it can be accepted by many medical units. The electronic medical record has simplified the of patient's information, as well as maintaining the use of the static image, already unable to meet the requirement of the pathology need. The related wireless networks can provide a certain bandwidth of a limited quality of the dynamic images. However, by using the compression technology of MPEG-4, our design can provide the electronic medical record at 24 frames per second for
receiving the video stream for Tablet PC. In addition, due to the performance difference of the CPU in PDA and Tablet PC, our prototype can present different performances as shown in Table 4.Because CPU operation ability institute of PDA is smaller than Tablet PC, so there will be a lot of differences in carrying out error rate,transmission frames and resolution Symposium on System Theory, 18-19 March 2002, Page(s): 406-411. [10] Morales, M.A.; Dalmiani, S.; Carpeggiani, C.; Macerata, A.; Ghione, S., Electronic medical records in a cardiological outpatient clinic, 2002 Proceedings of the Thirty-Fourth Southeastern Symposium on System Theory, 18-19 March 2002, Page(s): 381-384. Table 4 The performance measurements of the dynamic image of Tablet PC and PDA Picture (or frame) resolution Average image transmission rate Error rate Transmission frames per second Tablet 640x480 54 K bps 5.3% 24 fps PC PDA 320x240 37.7K bps 8.98% 6~8 fps References: [1] J. Presedo, D. Castro, J. Vila, M. Fednandez-Delgado, S. Fraga, M.LAMA, S. Barro, Wireless Interface for Monitored Patients in Coronary Care Units. Proceedings of the 22 nd Annual EMBS International Conference, July 23-28, 2000, Chicago IL, vol.3, Page(s): 1942-1945. [2] Ying-Wen Bai and Chien-Yung Cheng, Design and Implementation of a Remote Electrocardiogram System, 2003 IEEE International Symposium on Consumer Electronics, Dec. 2003, TA1-1, pp.14-19. [3] Jing Bai, Yonghong Zhang, and Bing Dai, Design and Development of an Interactive Medical Teleconsultation System over the World Wide Web, IEEE Transactions on Information Technology in Biomedicine, Volume: 2 Issue: 2, June 1998, Page(s): 74-79. [4] Wanjium Liao, Member, IEEE, Mobile Internet Telephony: Mobility Extension to H.323, IEEE Transactions on Vehicular Technology, Volume: 50 Issue: 6, Nov. 2001, Page(s): 1403-1414. [5] V. Parthasarathy, J. Modestino, and K. S. Vastola, Reliable transmission of high-quality video over ATM network, IEEE Trans. on Image Processing, Vol. 8, No.3, pp.361-374, March 1999. [6] Satyanarayanan, M, Accessing information on demand at any location. Mobile information access, IEEE Personal Communications, Volume: 3 Issue: 1, Feb. 1996, Page(s): 26 33. [7] Ching-Liang Su, Medical record and medical image processing by use the Internet SQL searching engine, 2000 Proceedings of International Symposium on Multimedia Software Engineering, 11-13 Dec. 2000 Page(s): 427-431. [8] Hall, E.S.; Vawdrey, D.K.; Knutson, C.D.; Archibald, J.K., Enabling remote access to personal electronic medical records. IEEE Engineering in Medicine and Biology Magazine, Volume: 22 Issue: 3, May-June 2003, Page(s): 133-139. [9] Dahle, J.; Callahan, D.W. Methodology for the development of an electronic medical record. 2002, Proceedings of the Thirty-Fourth Southeastern