《外文翻译-- Digital Demodulation in data acquisition system for multi-frequency electrical impedance tomography》由会员分享，可在线阅读，更多相关《外文翻译-- Digital Demodulation in data acquisition system for multi-frequency electrical impedance tomography（3页珍藏版）》请在金锄头文库上搜索。
1、Digital Demodulation in data acquisition system for multi-frequency electrical impedance tomography Shi Xuetao, You Fusheng, Ji Zhenyu, Fu Feng, Liu Ruigang, Dong Xiuzhen* School of Biomedical Engineering, Fourth Military Medical University, Xian, 710033, China. E-mail: AbstractIn order to extract
2、the impedance information at different frequencies in our multi-frequency electrical impedance tomography (EIT) system, a digital demodulation method was employed. Theoretically, this method can improve the SNR by 2/N times, if we sampling N point per period. Moreover, as very little system resource
3、s been needed in this method, all the calculating operations can be finished in a single Field Programmable Gate Arrays (FPGA) device instead of in a personal computer. And the final data acquisition system based on this method was established and the preliminary imaging results were obtained. Keywo
4、rds-electrical impedance system; data acquisition system; demodulation I. INTRODUCTION Biological tissue consists of a group of cells surrounded by tissue fluids. The cell can be idealized as an electrolyte containing a variety of sub-cellular structures that was enveloped completely by a membrane w
5、hich has a low electrical leakage in the resting state. The tissue fluids also can be seen as an electrolyte. Therefore, each tissues electrical property will change as the probe frequency varies. The impedance of living tissue at each frequency can be calculated by following equation 1: )/2(10cffjR
6、RRZ+= (1) Where R0 is the low frequency resistance of living tissue, R is the high frequency resistance, is dispersion parameter, and fc is the characteristic frequency. For different tissue or tissue at different condition, these parameters will be different. In our multi-frequency electrical imped
7、ance tomography (MFEIT) system, the tissue impedance is detected at least in two separate frequencies simultaneously, and the resulted image represents the difference of impedance at these frequencies. As the impedance information was modulated by stimulation signal that was injected into the target
8、 area through the boundary electrodes. The way to extract it is a key problem in the data acquisition system. Some systems employ FFT methods. But this is not a best solution for real-time system because it will bring about the problem of mass data transportation and low calculation speed. In order
9、to get a better solution, a digital demodulation method based on property of triangle function was employed. II. PRINCIPLE AND PERFORMANCES A. Principle The stimulation signal (current) in our system can be described by equation: =10)sin()(nllltkAtS (2) Where Al is amplitude of stimulation signal at
10、 each frequency, kl is angel frequency, and n is the total number of frequencies. When it was injected into target area, and sampling N (N2maxk0, k1, , kn-1) points per period, the following sequence is got: =+=10)2sin()(nllllliNkZAiSn Where Zl is the tissues impedance modular at each frequency, l i
11、s phase shift, i is an integer from 0 to N-1, n is the frequency number. The Zl and the l are the parameters we needed. Considering the orthogonal property of the triangle function, we made the reference sine sequence as: 2/)2cos(2)(; 2/)2sin(2)(21ikNAiSikNAiSllll= Then, the inner product between Sn
12、 and S1 and between Sn and S2 will be: )sin( ),cos(21llllllZPZP= It can be see that Pl1 and Pl2 are exactly the real part and imaging part of the tissues impedance. Therefore, the Zl and l can be calculated easily as: Xiuzhen Dong is the corresponding author (phone & fax: 86-29-84776397; e-mail: ).
13、 978-1-4244-4713-8/10/$25.00 2010 IEEE)arctan( ,122221llllllPPPPZ=+= B. The Anti-noise Ability of Digital Demodulator Suppose that a Gaussian white noise n, with an average of 0 and variance of 2, was interfered into the measured signal. The signal to noise ratio (SNR) at each frequency will be: )si
14、n(;)cos(21llllllllZASNRZASNR= After demodulation, the result will be: )()()2cos(2);()()2sin(220221011inPinNikNAAPinPinNikNAAPlNilllllNillll +=+=+=+= According to the features of Gaussian white noise, the n and n should also be as a Gaussian white noise. And its variance should be: 220222022)(2)2(cos
15、)(;2)(2)2(sin)(llNilllNilNANAinDNiknDNANAinDNiknD= = Thus, after demodulation the SNR of Pl1 and Pl2 will be: 221122)sin(22)cos(llllllllllSNRNZANSNRSNRNZANSNR= It can be seen that the SNR was improved by 2/Ntimes. III. SYSTEM IMPLEMENTATION As been shown in figure 1, the final data acquisition syste
16、m was consists of a programmable current source, a programmable gain voltage meter, a multiplexer, a 16-bit high speed ADC, a USB interface and a personal computer. The current source, the ADC and the voltage meter are all controlled by a FPGA device. The USB interface receives the operation command
17、s the computer made and send them to FPGA. In FPGA, those commands were translated and been sent to proper part. The current source was based on digital synthesize technique. In this technique, a series of driving signals sampled data was stored in the memory units of the FPGA previously, and was re
18、ad out and been sent to a high speed 16-bit DAC periodically. After smoothed by a lower pass filter, the reconstructed signal was converted into a driving current by a voltage control current source which was reported in 2. TargetProgrammablecurrent source Programmablegain voltage meterMuxADCFPGAUSB
19、interfaceelectrode Figure 1. Block diagram of data acquisition system During data acquisition period, the FPGA will configure the multiplexer to connect the stimulation signal to proper electrode, and a pair of adjacent electrode was connected to the voltage meter. After been amplified, the voltage
20、difference between these electrode pair was converted into digital signal with 128 points per signal periods by the following ADC. The resulted digital signal was store into a FIFO inside the FPGA. Previously, those digital signals were transferred to computer through the USB interface and the final
21、 demodulation operation was conducted by the computer. Now all the demodulation operation was conducted by the FPGA itself. The realization of those demodulation operations can be illustrated by figure 2. In this part, the reference sequence was stored into a memory unit inside the FPGA previously.
22、After the voltage signal been sampled and digitalized, it was read out sequentially from the FIFO. At the same time, the reference sequence was also read out. Both of these data were sending to a 16-bit multiplier, and the product was send to an accumulator. After totally 128 times operation finishe
23、d, the demodulation result was produced and stored in accumulator. IV. SYSTEM PERFORMANCE The final data acquisition system can collect the impedance information at four frequencies. When configured as a 16-electrode system, the maximum data acquisition speed is 8 frames per second. At present bandw
24、idth, which is from 1kHz to 200kHz, the measured precision at a 20 is better than 70dB while the amplitude of the driving current is 0.5mA. This can meet the accuracy demand that 3 suggested. The demodulated results can be seeing in figure 3 where synchronization sample demodulation results 4 were a
25、lso obtained and shown in broken line. From the figure it can be see that the final data acquisition precision was improved by at lest 10dB compared with the synchronization sample method. V. SALINE TANK IMAGING RESULTS After the system been established, a circular Lucite tank with a height of 10cm
26、and 30cm in diameter, and filled with 4000g of 0.075% saline solution was used in the imaging experiment. There are 16 Ag/AgCl electrodes 10mm in diameter equally spaced on the tank periphery. A finger of FIFORAMMultiplierAccumulatorCLKAddressgenerator Figure 2. Block diagram of digital demodulation
27、 Figure 3. Demodulated result as a function of measuring frequency one of our colleagues was immersed in the saline tank and move from peripheral area to central area, and a beaker 4cm in diameter was placed in the central of the tank. The voltage differences at 25kHz and 50kHz were acquired synchro
28、nously. The modulus of voltage differences at 25kHz were selected as the reference data sets, and that at 50kHz were used as target data sets. The reconstruction algorithm used in this paper is a weighted equipotential backprojection algorithm 5. By this algorithm, a region with resistivity decrease
29、d is indicated by a dark area in corresponding area in images. Figure 4 shows the resulted images while the finger is 1cm and 3cm from the fringe of the tank. It can be see that though the resistivity of beaker is much higher than other area inside the tank, only the position of the finger was refle
30、cted in images. This is caused by the fact that only the fingers resistivity changed with the frequency. VI. DISCUSSIONS AND CONCLUSIONS The demodulation method in multi-frequency EIT data acquisition system has a strong influence on the precision of sampled data. According to the analysis above, th
31、e digital demodulation method in this paper can diminish the random noise greatly. As in most cases, the systems SNR are determined mainly by the root mean square (RMS) value of random noise, the system performance can be improved greatly by this method. From the theoretical analysis above, it can a
32、lso be see that the anti-noise ability of the digital demodulation method was mainly determined by the number of points that we sampled during the boundary voltage signal circle. The more point we sampled, the more noise been rejected in final results. However, increase the number of samples, may al
33、so bring about the disadvantages of more data to be processed and much fast ADC to be chosen. In order to overcome the first problem, we managed the FPGA in system to handle the all calculations, and only the final results been retained and transferred to the computer to reconstruct the EIT images.
34、As for the second problem, for the period of maximum frequency signal in this system is 5s, it is unadvisable select a very high speed ADC to sample 128 points within 5s. Therefore, an ADC with a maximum sample rate of 1MSPS was selected. While the signal frequency is higher than 7.8kHz, the samplin
35、g was conducted in more than one signal circle to ensure all of the 128 points been sampled. Through the methods described above, the final system exhibited a high precision property. The imaging results based on saline tank also demonstrated that the system can be used to get multi-frequency EIT im
36、ages. Figure 4. Quasi-static images of physical phantom with a finger which is 1cm (left), and 3cm (right) from the fringe of the tank. ACKNOWLEDGMENT This work was supported in part by the National Natural Science Foundation of the Peoples Republic of China under Grant 50337020. REFERENCES 1 K. Col
37、e and R Cole, Dispersion and absorption in dielectrics. I. Alternating current characteristics, J. Chem. Phys. 1941, 9, pp 341-351. 2 R Brags, Jr Rosell and P Riu, A wide-band AC-coupled current source for electrical impedance tomography, Physiol. Meas., 1994, 15 (Suppl.), pp A91A99. 3 P. Riu, J. Ro
38、sell, A. Lozano et al. Multifrequency static imaging in electrical impedance tomography: Part 1 instrumentation requirements. Med. Biol. Eng. Comp. 1995, 33, pp 78492. 4 R Halter, A Hartov and K Paulsen, Design and implementation of a high frequency electrical impedance tomography system, Physiol. Meas. 2004, 25 (1) , pp 379390. 5 N. Avis and D. Barber, Incorporating a-priori Information into the Sheffield Filtered Backprojection Algorithm. Physiol. Meas. 1995, 16 (Suppl. 3A) pp 111-122. Introduction (Heading 1)
