GASTROESOPHAGEAL reflux disease (GERD) is one of

Transcription

1 1692 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 52, NO. 10, OCTOBER 2005 Application of the Empirical Mode Decomposition to the Analysis of Esophageal Manometric Data in Gastroesophageal Reflux Disease Hualou Liang*, Senior Member, IEEE, Qiu-Hua Lin, and J. D. Z. Chen, Senior Member, IEEE Abstract The Empirical Mode Decomposition (EMD) is a general signal processing method for analyzing nonlinear and nonstationary time series. The central idea of EMD is to decompose a time series into a finite and often small number of intrinsic mode functions (IMFs). An IMF is defined as any function having the number of extrema and the number of zero-crossings equal (or differing at most by one), and also having symmetric envelopes defined by the local minima, and maxima respectively. The decomposition procedure is adaptive, data-driven, therefore, highly efficient. In this contribution, we applied the idea of EMD to develop strategies to automatically identify the relevant IMFs that contribute to the slow-varying trend in the data, and presented its application on the analysis of esophageal manometric time series in gastroesophageal reflux disease. The results from both extensive simulations and real data show that the EMD may prove to be a vital technique for the analysis of esophageal manometric data. Index Terms Empirical mode decomposition, esophageal motility, gastroesophageal reflux disease, lower esophageal sphincter. I. INTRODUCTION GASTROESOPHAGEAL reflux disease (GERD) is one of the most prevalent gastrointestinal diseases [1]. It is characterized by excessive reflux of gastric content (acid, pepsin, etc.) into the esophagus causing symptoms (heartburn, acid regurgitation, etc.) and mucosal inflammation and injuries. GERD occurs when the lower esophageal sphincter (LES) is of low pressure and/or relaxes inadequately, and stomach contents leak back, or reflux, into the esophagus. The LES is a ring of muscle at the bottom of the esophagus that acts like a valve between the esophagus and stomach. Therefore, LES functions as a barrier, preventing the reflux of gastric content into the esophagus. When the LES is weak, the powerful acid which helps digest the food can reflux back into the esophagus causing inflammation and pain. There is a positive pressure gradient between the stomach and the esophagus that tends to promote the reflux from the stomach into the esophagus. LES normally can counteract this Manuscript received September 8, 2004; revised December 23, Asterisk indicates corresponding author. *H. Liang is with the School of Health Information Sciences, University of Texas at Houston, 7000 Fannin, Suite 600, Houston, TX USA ( hualou.liang@uth.tmc.edu). Q.-H. Lin is with the School of Electronic and Information Engineering, Dalian University of Technology, Dalian , P. R. China. J. D. Z. Chen is with the Division of Gastroenterology, University of Texas Medical Branch, Galveston, TX USA. Digital Object Identifier /TBME pressure gradient and effectively prevent reflux by maintenance of a relatively higher resting pressure. Low resting LES pressure usually leads to episodes of gastroesophageal reflux. This has been frequently observed in patients with severe GERD [2], [3]. The use of the LES pressure to help diagnosing GERD, although controversial [4], has nevertheless been considered as a potential predictor of erosive esophagitis [5]. The purpose of esophageal manometry is to measure LES pressure and evaluate esophageal contractions. The station pullthrough method is usually performed to measure the location, length and resting pressure of the LES as well as its pressure during swallows. The normal range of the LES resting pressure is mmhg. Fig. 1 shows a typical recording of the esophageal manometric data. The elevated pressure with sharp rising edge is the LES pressure which is extremely important in the diagnosis of GERD. The signal is highly nonstationary due to the nature of the station pull-through operation. The respiratory contamination is another added variation which poses a major challenge to the analysis of LES pressure. At the trough, middle and valley of the respiration are all used in current clinical practice to estimate the LES pressure, but none is accurate for the assessment of esophageal dismotility. It is the subject of this paper to accurately measure LES pressure and evaluate esophageal contractions. In this paper, we present a general nonlinear, nonstationary data analysis method empirical mode decomposition (EMD)[6] for the analysis of esophageal manometric data. An automatic procedure is developed for the extraction of the LES pressure based on the statistical properties of the IMFs derived from the EMD. Through extensive computer simulations and real data, we show that the EMD is indeed capable of removing the considerable noise contaminated the LES pressure signal. II. METHODS A. The Measurement of LES A thin, flexible, plastic tube called a manometry catheter with four solid-state pressure sensors along its wall was gently passed through the nose, down the back of the throat, and into the esophagus. The distance between adjacent pairs of sensors was 5 cm. Once the catheter was in place, the subject was instructed to lay supine on his back with his arms and hands at his sides, and the catheter was taped to his nose to prevent movement. The tube inside the esophagus allows the pressures generated by the esophageal muscle to be measured /$ IEEE

2 LIANG et al.: APPLICATION OF THE EMD TO THE ANALYSIS OF ESOPHAGEAL MANOMETRIC DATA 1693 Fig. 1. A 10-min recording of the esophageal manometric data. The elevated pressure with sharp rising edge is the LES pressure as labeled with a horizontal bar. during rest or swallows. The procedure was performed using the pull-through method. Initially, all sensors are located in the stomach and then the catheter is pulled up 1-cm at the time, a high pressure is measured when a sensor enters the lower esophageal sphincter. In some cases, the sensor is pulled through the sphincter several times to ensure an accurate measure of the LES pressure. The procedure takes approximately 45 min to 1 h and the patient should be fasting for 4 6 h before the study. The data is sampled at 4 Hz. B. Empirical Mode Decomposition (EMD) EMD is a general nonlinear, nonstationary signal processing method. The EMD method was initially proposed for the study of ocean waves [6], and found immediate applications in biomedical engineering [7], [8]. The major advantage of EMD is that the basis functions are derived directly from the signal itself. Hence, the analysis is adaptive, in contrast to Fourier analysis, where the basis functions are linear combinations of fixed sinusoids. The principle of EMD is to decompose a signal into a sum of oscillatory functions, namely intrinsic mode functions (IMFs), that: 1) have the same numbers of extrema and zero-crossings or differ at most by one; and 2) are symmetric with respect to local zero mean. With these two requirements, the meaningfully instantaneous frequency of an IMF can be well defined. Otherwise, if blindly applied to any signal, the instantaneous frequency may result in a few paradoxes [9], [10]: it may go beyond the band for bandlimited signal or it may not represent one of the frequencies in the Fourier spectrum in the global sense. As such, the two conditions of an IMF allow the calculation of a meaningfully instantaneous frequency. Specifically, the first condition is similar to the narrow-band requirement, whereas the second condition modifies a global requirement to a local Fig. 2. A toy example of EMD Decomposition. Left column: a Gaussian amplitude-modulated linear chirp (top), a triangular waveform (middle) and their composite signal (bottom). Right column (from top to bottom): two components (IMFs) extracted by EMD revealing a striking agreement with the signals shown in the left column, as well as the last final residue. one by using the local mean of the envelopes defined by the local maxima and the local minima, and is necessary to ensure that the instantaneous frequency will not have unwanted fluctuations as induced by asymmetric waveforms. To make use of EMD, the signal must have at least two extrema one maximum and one minimum to be successfully decomposed into IMFs. Given these two definitive requirements of an IMF, the sifting process for extracting IMFs from a given signal, is described as follows. 1) Identify all the maxima and minima of. 2) Generate its upper and lower envelopes, and, with cubic spline interpolation.

3 1694 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 52, NO. 10, OCTOBER 2005 Fig. 3. The simulation example of extraction of the LES pressure signal using the EMD and the lowpass filtering. (a) A segment simulated LES pressure signal. (b) Respiratory artifacts obtained from a sensor in the stomach. (c) Artificially generated Gaussian noise (see text for details). (d) The synthesized composite signal using (a) (c) in thin line, and the extracted pressure signal using EMD in thick line. (e) The same as (d) except the thick curve obtained by the low-pass filtering. Fig. 4. Ten IMF components and the residual (C11 on the bottom) of simulated LES pressure signal in Fig. 3(d) obtained by EMD method. 3) Calculate the point-by-point mean from upper and lower envelopes,. 4) Extract the detail,. 5) Check the properties of : if meets the above-defined two conditions, an IMF is derived and replace with the residual ; If is not an IMF, replace with. 6) Repeat Steps 1) 5) until the residual satisfies some stopping criterion. At the end of this process, the signal follows: can be expressed as where is the number of IMFs, denotes the final residue which can be interpreted as the dc component of the signal, and are nearly orthogonal to each other, and all have nearly zero means. Due to this iterative procedure, none of the sifted IMFs is derived in closed analytical form.

4 LIANG et al.: APPLICATION OF THE EMD TO THE ANALYSIS OF ESOPHAGEAL MANOMETRIC DATA 1695 Fig. 5. Illustration of the EMD acting as a low-pass filter through the reconstruction of the data from the IMF components. In practice, after a certain number of iterations, the resulting signals do not carry significant physical information, because, if sifting is carried on to an extreme, it could result in a pure frequency modulated signal of constant amplitude. To avoid this we can stop the sifting process by limiting the normalized standard deviation, computed from two consecutive sifting results. The is defined as The is usually set between 0.2 and 0.3. By construction, the number of extrema is decreased when going from one residual to the next, and the whole decomposition is guaranteed to be completed with a finite number of modes. Fig. 2 shows a simulated example of EMD decomposition, where the analyzed signal (bottom left) is composed of a Gaussian amplitude-modulated linear chirp (top left) and a triangular waveform (middle, left). The EMD, when applied to the signal, yields two IMF components and the final residual shown in Fig. 2 (right column). These two IMFs bear a striking similarity to the signals shown in Fig. 2 (left column). With the presence of the nonharmonic triangular waveform, any harmonic analysis such as Fourier transform would end up with a much less compact and physically less meaningful decomposition [11]. C. Statistical Identification of Trend By the nature of the decomposition procedure, the data is decomposed into fundamental components, each with distinct time scale. More specifically, the first component as the smallest time scale which corresponds to the fastest time variation of data. As the decomposition process proceeds, the time scale increases, and hence, the mean frequency of the mode decreases. Based on this observation and the above equation, we may devise a general purpose time-space filtering as where,. For example, when and, it is a high-pass filtered signal; when and, it is a low-pass filtered signal; when, itisa bandpass filtered signal. The above equation forms the basis for our application of esophageal manometric data described below, where we use it as a low-pass filtering. The use of the EMD as a filter is essentially a partial reconstruction process, either from fine-to- coarse (i.e., high-pass filtering by adding fast oscillations up to slow oscillations), from coarse-to-fine (i.e., low-pass filtering), or a collection of intermediate IMF oscillation components (i.e., bandpass filtering). In case of the fine-to-coarse partial reconstruction, for example, the filtered signal can be specifically written as: where, is the total number of IMFs in the data. The only open question for utilizing the above equation as a filter is how to choose the parameter. We tackle this question by taking advantage of IMFs statistics that each has zero-mean. As such, we design a three-step procedure to identify the slow-varying trend: 1) the mean and the standard deviation of taken over time are performed as a function of ; 2) one sample -test is employed to determine when the mean significantly departs from zero; and 3) once is identified as a significant change point, partial reconstruction with IMFs from up to the residual renders us the slow-varying trend in the

5 1696 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 52, NO. 10, OCTOBER 2005 Fig. 6. The mean and its standard deviation of the fine-to-coarse reconstruction as a function of index K. The vertical dash-line at K =6indicating that the mean departs significantly from zero (p <0:01). The filtered pressure signal partially reconstructed from IMFs 7 11 is shown in Fig. 3(d) (thick line). Fig. 7. Power spectra of composite signal (thick line) shown in Fig. 3(d) in thin line and of the data obtained by summing the faster time-scale IMFs from C1 to C6 (thin line). The vertical dash-line at 0.12Hz indicating the cut-off frequency used in the low-pass filtering and the selection of six components here is due to Fig. 6. data. In what follows, we apply this simple idea to characterize the LES pressure. III. RESULTS A. Simulations Computer simulations were conducted to validate the performance of the EMD method. The simulated esophageal manometric signal [Fig. 3(d), thin line] was composed of a simulated pressure signal [Fig. 3(a)], actual respiratory artifacts [Fig. 3(b)] obtained from a sensor in the stomach, and simulated Gaussian noise of different noise level proportional to the strength of the pressure signal [Fig. 3(c)]. The pressure signal is divided into five quantile-based levels and contaminated by Gaussian noise with five different noise variance ranging from 1.0 to 3. The EMD method yields ten IMF components together with the final residual as shown in Fig. 4. To illustrate how the EMD can be used as a low-pass filter, we reconstituted the original data from the IMF components.

6 LIANG et al.: APPLICATION OF THE EMD TO THE ANALYSIS OF ESOPHAGEAL MANOMETRIC DATA 1697 Fig. 8. The IMFs for the data show in Fig. 1 through the EMD method. Fig. 9. The mean and its standard deviation of the fine-to-coarse reconstruction as a function of index K. The vertical dash-line at K =7indicating that the mean departs significantly from zero (p <0:01). The step-by-step reconstruction is shown in Fig. 5 where the original data is plotted in gray lines and partial sum of the IMFs in thick lines. The very first plot shows the data and the last component c11, the residue of the sifting, which denotes the dc component in the data. The very last plot shows the summation of all the IMFs, which looks like the original data. The intermediate plots show the progress of addition of the IMF components. If we stopped at any step, the data was filtered. Fig. 6 shows the evolution of the mean (the curve below the zero-line) and its standard deviation (the error bars) of the fine-to-coarse partial reconstruction as a function of, with indicating that the mean departs significantly from zero. The filtered pressure signal partially reconstructed from IMFs 7 11 is shown in Fig. 3(d) in thick line, where we can see clearly that the EMD is able to track the pressure signal very well. To quantify the performance of the EMD method, we used the root-mean-square error (RMSE) as the error measure where is the filtered pressure signal [Fig. 3(d), thick line], is the original pressure signal [Fig. 3(a)], and is the number of time points. The RMSE is , which is believed to be negligible. Comparison with conventional low-pass filtering is shown in Fig. 3(e), where the low-pass filtered data (thick line) is

7 1698 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 52, NO. 10, OCTOBER 2005 Fig. 10. Extraction of the LES pressure signal using the EMD. The original recording is shown in thin line and the filtered pressure signal (thick line) is obtained from the partial reconstruction with IMFs 8 to 11 shown in Fig. 8. superimposed on the simulated esophageal manometric signal (thin line). The selection of the cut-off frequency in the low-pass filtering (indicated by vertical dash-line at 0.12Hz in Fig. 7) is determined by comparison of power spectra of composite signal with that of data obtained by summing the faster time-scale IMFs from C1 to C6, where these six IMFs, established in Fig. 6, represent the noise contamination in the simulated composite signal. It is evident that the sharp edges of the LES pressure are not preserved by the low-pass filter method, which is otherwise important for accurate measurement of the duration of the LES pressure. The RMSE is Indeed, the EMD performed better than the low-pass filter method. B. Applications to Esophageal Manometric Data The experimental data described previously were used in this study. As our first example, we applied the EMD to the data shown in Fig. 1. The EMD method yielded ten IMF components as well as the residual as shown in Fig. 8. The filtered pressure signal (by summing the last 4 components from C8 to C11; the selection of the 4 components was determined by our 3-step statistical procedure presented in Fig. 9 with as the change point), shown in Fig. 10 in thick line, is superimposed on the data in thin line. To further illustrate the effectiveness of the EMD method, two more applications are presented in Fig. 11(a) and (b), respectively. In each example, data are shown as thin lines and the extracted LES pressure signal as thick lines. As we can see from both examples, the EMD performs equally well as previous example regardless of the high nonstationarity of the data. Our final example is to illustrate the potential of the EMD method that could be used for the purpose of the diagnosis. In this example, a segment of the recording (Fig. 12) was taken before, during and after a swallow of 5 ml water, a clinical procedure to assess the relaxation of the lower esophageal sphincter. The relaxation of the LES is considered as normal if the residual pressure is below 5 mmhg and as abnormal if the residual pressure is above 8 mmhg during a wet swallow. From the Fig. 12, we see that, just after 1.5 min, there is a deep drop of the LES pressure; this relaxation is brief but extremely important in diagnosing patients with swallowing disorders, such as achalasia (Achalasia is a motor disorder of the esophagus characterized by complete loss of the LES relaxation during swallowing). We see that, from the Fig. 12, both the EMD method and the low-pass filtering (its cut-off frequency is determined by the procedure as described in the Simulations) can be used to reduce the noise contamination of the data. However, the EMD method is clearly able to track the brief, sharp drop of the LES pressure (shown thick, solid line in Fig. 12), whereas the low-pass filtering smoothens out the brief relaxation, which would unavoidably reduce the accuracy in the assessment of the LES relaxation. IV. DISCUSSION AND CONCLUSION EMD is an emerging new technique for adaptively decomposing nonstationary signal in a sum of local oscillatory components (IMFs). It is local in time, fully data-driven, and does not require any prior knowledge on the nature and the number of IMF components embedded in the data. This technique has already been applied with success in biology and medicine [7], [8], [12], [13]. In this contribution, we applied the idea of EMD to develop strategies to automatically identify the relevant IMFs that contribute to the slow-varying trend in the data, and presented its application on the extraction of the LES pressure in

8 LIANG et al.: APPLICATION OF THE EMD TO THE ANALYSIS OF ESOPHAGEAL MANOMETRIC DATA 1699 Fig. 11. (a) and (b) Two more application examples to extract the LES pressure signal using the EMD. In each panel, the original recording is shown as a thin line, and the extracted LES pressure signal as a thick red line. GERD. We showed through both computer simulations and actual data, that the method is able to successfully extract the LES pressure signal and compares favorably to the conventional low-pass filtering. In the procedure of statistical identification of signal trend, the t-test is used to determine the significant change point. The use of the t-test assumes that the samples are drawn from a random normal distribution. We meet this assumption by using a large number of samples, the high significant level and symmetric sample distributions in the study. In addition, the procedure works efficiently for the data in which it contains a slowvarying trend. The decomposition technique is based on the local characteristic time scale of the data, whose basis functions (or IMFs) used to represent the given signal are nonlinear functions that are directly extracted from the data. Therefore, the time scale is defined by the data per se, rather than by a pre-determined value. Fourier analysis cannot separate these IMFs without using preassigned cut-off frequencies. This is the crucial difference between EMD and Fourier-based filtering. The use of the EMD

9 1700 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 52, NO. 10, OCTOBER 2005 Fig. 12. A 3-min-long recording was taken before, during and after a swallow of 5 ml water, a clinical procedure to assess the relaxation of the lower esophageal sphincter. The LES pressure signal extracted by the EMD (solid, fat line) and that by the conventional low-pass filter (dashed line) are overlaid with the raw recording (solid, thin line). We see, with the low-pass filtering, that the accurate assessment of the LES relaxation becomes impossible. But, the EMD method is able to track the brief relaxation of the LES pressure very well. process as a filter and its comparisons with wavelet and Fourier analyses have just recently been studied [14], [15]. The EMD, although quite simple and highly effective, still lacks solid theoretical foundations. The application of this method also usually requires additional steps to be taken to postprocess the output of EMD. For example, Hilbert transform can be used to characterize the instantaneous frequencies of IMFs [5]. In addition, the implementation of EMD requires some attention to deal with the management of the end points for cubic splines interpolation in the EMD process and the selection of the stopping criteria for the sifting. The stopping criteria, however, are not necessarily unique; the EMD is robust to various adjustable parameters in the sifting process [16]. All in all, we have introduced here a new method to successfully extract the LES pressure in GERD. ACKNOWLEDGMENT The authors would like to thank the reviewers for valuable comments to improve the paper. REFERENCES [1] R. S. Sandler, J. E. Everhart, M. Donowitz, E. Adams, K. Cronin, C. Goodman, E. Gemmen, S. Shah, A. Avdic, and R. Rubin, The burden of selected digestive diseases in the United States, Gastroenterology, vol. 122, pp , [2] J. Behar, P. Biancani, and D. G. Sheahan, Evaluation of esophageal tests in the diagnosis of reflux esophagitis, Gastroenterology, vol. 71, pp. 9 15, [3] P. J. Kahrilas et al., Esophageal peristaltic dysfunction in peptic esophagitis, Gastroenterology, vol. 91, pp , [4] P. J. Kahrilas, R. E. Clouse, and W. J. Hogan, American-gastroenterological-association technical review on the clinical use of esophageal manometry, Gastroenterology, vol. 107, pp , [5] M. P. Jones, S. S. Sloan, J. C. Rabine, C. C. Ebert, C. F. Huang, and P. J. Kahrilas, Hiatal hernia size is the dominant determinant of esophagitis presence and severity in gastroesophageal reflux disease, Am. J. Gastroenterol., vol. 96, pp , [6] N. E. Huang, Z. Shen, S. R. Long, M. C. Wu, H. H. Shih, Q. Zheng, N.-C. Yen, C. C. Tung, and H. H. Liu, The empirical mode decomposition and the Hilbert spectrum for nonlinear and nonstationary time series analysis, Proc. Roy. Soc. Lond., vol. A 454, pp , [7] W. Huang, Z. Shen, N. E. Huang, and Y. C. Fung, Engineering analysis of biological variables: An example of blood pressure over 1 day, Proc. Nat. Acad. Sci. USA., vol. 95, pp , [8] H. Liang, Z. Lin, and R. W. McCallum, Artifact reduction in electrogastrogram based on the empirical model decomposition method, Med. Biol. Eng. Comput., vol. 38, pp , [9] B. Boashash, Estimating and interpreting the instantaneous frequency of a signal-part I: Fundamentals, Proc. IEEE, vol. 80, pp , [10] L. Cohen, Time-Frequency Analysis. Englewood Cliffs, NJ: Prentice- Hall, [11] G. Rilling, P. Flandrin, and P. Goncalves, On empirical mode decomposition and its algorithms, in Proc. IEEE-EURASIP Workshop on Nonlinear Signal and Image Processing, 2003, Grado (I). [12] R. Balocchi, D. Menicucci, E. Santarcangelo, L. Sebastiani, A. Gemignani, B. Ghelarducci, and M. Varanini, Deriving the respiratory sinus arrhythmia from the heartbeat time series using empirical mode decomposition, Chaos, Solitons, Fractals, vol. 20, pp , [13] H. Liang, S. L. Bressler, R. Desimone, and P. Fries, Empirical mode decompositionof local field potentials from macque V4 in visual spatial attention, Biol. Cybern., vol. 92, pp , [14] Z. Wu and N. E. Huang, A study of the characteristics of white noise using the empirical mode decomposition method, Proc. Roy. Soc. London A, vol. 460, pp , [15] P. Flandrin, G. Rilling, and P. Goncalves, Empirical mode decomposition as a filter bank, IEEE Sig. Process. Lett, vol. 11, no. 2, pp , Feb [16] N. E. Huang, M.-L. C. Wu, S. R. Long, S. S. P. Shen, W. Qu, P. Gloersen, and K. L. Fan, A confidence limit for the empirical mode decomposition and Hilbert spectral analysis, Proc. Roy. Soc. London A, vol. 459, pp , 2003.

10 LIANG et al.: APPLICATION OF THE EMD TO THE ANALYSIS OF ESOPHAGEAL MANOMETRIC DATA 1701 Hualou Liang (M 00 SM 01) received the M.Sc. in electronic engineering from Dalian University of Technology, Dalian, China, and Ph.D. in Physics from Chinese Academy of Sciences, Beijing, China. He is an Assistant Professor in University of Texas Health Science Center, Houston, TX. He was postdoctoral researcher in Tel-Aviv University, Tel-Aviv, Israel, Max-Planck-Institute for Biological Cybernetics, Tuebingen, Germany, and the Center for Complex Systems and Brain Sciences, Florida Atlantic University, Boca Raton. He has written more than 50 papers, conference proceedings and book chapters. His research experience throughout the years ranges from areas in biomedical signal processing to cognitive and computational neuroscience. He teaches graduate interdisciplinary courses including Biomedical Signal Processing and Computational Biomedicine. Qiu-Hua Lin received the B.Sc. degree in wireless communication, the M.Sc. degree in electronics engineering, and the Ph.D. in signal and information processing, all from Dalian University of Technology (DUT), Dalian, China. She is currently an Associate Professor at the School of Electronic and Information Engineering, DUT, Dalian, China. Her research interests include blind signal processing, array signal processing, biomedical signal processing, and secure communication. J. D. Z. Chen (S 88 M 88 SM 94) received his predoctoral training in China and his graduate education at the Catholic University of Leuven, Leuven, Belgium. While there he started his research on gastrointestinal motility with a group of internationally renowned investigators in the Department of Medicine. He received the Ph.D. degree from the same university in He moved to the United State and accepted a faculty position at the University of Virginia School of Medicine, Charlottesville. In June 1999, he joined the Division of Gastroenterology, Department of Internal Medicine at The University of Texas Medical Branch, Galveston, where he is currently an Associate Professor and Director of Clinical Physiology Laboratory. His current research interest is in the area of physiology and pathophysiology of gastrointestinal motility. His specific research topics include clinical gastrointestinal motility studies, electrogastrography, gastrointestinal pacing in animal models and patients with severe motility disorders, autonomic functions and biomedical signal processing. His research during the past 5 years has been supported by over 20 grants from the federal government, state, private foundations and industries. Dr. Chen serves as the first President of the International Electrogastrography Society, has given over 130 invited lecture in various countries, served on committees of national and international conferences, published one book, over 160 peer-reviewed journal articles and book chapters, and over 300 conference papers and abstracts. He is a fellow of the American College of Gastroenterology and a member of various professional societies in medicine and biomedical engineering.