Advertisement
< Vol. 37 Issue 5 | May 2014 | pp. 349 - 426

Assessment and in vitro experiment of artificial anal sphincter system based on rebuilding the rectal sensation function

Abstract

In this paper, a novel artificial anal sphincter (AAS) system based on rebuilding the rectal sensation function is proposed to treat human fecal incontinence. The executive mechanism of the traditional AAS system was redesigned and integrated for a simpler structure and better durability. The novel executive mechanism uses a sandwich structure to simulate the basic function of the natural human anal sphincter. To rebuild the lost rectal sensation function caused by fecal incontinence, we propose a novel method for rebuilding the rectal sensation function based on an Optimal Wavelet Packet Basis (OWPB) using the Davies-Bouldin (DB) index and a support vector machine (SVM). OWPB using a DB index is used for feature vector extraction, while a SVM is adopted for pattern recognition.

Furthermore, an in vitro experiment with the AAS system based on rectal sensation function rebuilding was carried out. Experimental results indicate that the novel executive mechanism can simulate the basic function of the natural human anal sphincter, and the proposed method is quite effective for rebuilding rectal sensation in patients.

Int J Artif Organs 2014; 37(5): 392 - 401

Article Type: ORIGINAL RESEARCH ARTICLE

DOI:10.5301/ijao.5000308

Authors

Peng Zan, Jinding Liu, Enyu Jiang, Hua Wang

Article History

  • • Accepted on 10/01/2014
  • • Available online on 01/03/2014
  • • Published in print on 15/06/2014

Disclosures

Financial Support: This work was supported by National Natural Science Foundation of China (No. 31100708, No. 31370998) and Natural Science Foundation of Shanghai (No.11ZR1412400).
Conflict of Interest: None.

This article is available as full text PDF.

Download any of the following attachments:

INTRODUCTION

Fecal incontinence is the result of numerous disorders affecting the anatomy and physiology of the anorectum. It mainly manifests as recurrent, uncontrolled passage of solid or liquid dejecta or gas. Fecal incontinence not only brings physical and mental sufferings to patients, but also seriously affects patients’ orthobiosis and work (1). Traditional therapeutic methods such as biofeedback (2, 3), sacral nerve stimulation (4, 5), colostomy, and ileostomy (6, 7) can treat fecal incontinence to some degree, however, with limited success.

An artificial anal sphincter (AAS) has been employed to alleviate fecal incontinence. The traditional structure of an AAS is composed of a circular compression cuff enclosing the anal canal, a pressure-regulating balloon, and a separate pump by which the patient can inflate and deflate the cuff (8, 9). However, there are mainly two limitations for the existing AAS. One is that patients have to pinch the pump with their hand in order to defecate. The other defect is that patients are unaware of when to defecate because of the loss of rectal sensation function, which has become the most sophisticated problem restricting the development of an AAS. Thus, there is an urgent need to rebuild rectal sensation function in order to promote the further development of an AAS system.

To solve these problems, a new AAS prosthesis based on rectal sensation function rebuilding is presented in this paper. The aim is to redesign and integrate a novel executive mechanism with a simpler structure and better durability to simulate the normal physiology of the human anorectum. The detailed structure model of the executive mechanism is discussed, and then its performance characteristics are assessed with a measurement experiment we established to ensure reliability.

Furthermore, a method for rectal sensation function rebuilding based on optimal wavelet packet (OWP) and support vector machine (SVM) is proposed to predict the urge to defecate. Finally, an in vitro experiment of the novel AAS system based on rectal sensation function rebuilding is carried out. Experimental result indicates that the proposed method is quite effective to predict the urge to defecate.

SYSTEM OVERVIEW

The novel AAS system proposed in this paper is based on rectal sensation function rebuilding to treat fecal incontinence. The system consists of two components: the one, implanted in the body of patients, is called the implanted component, while the other, set outside the body, are the outside components. The two components are highly integrated and the fact that there is no wire linking the implanted and outside components ensures that the implantation is less risky and has fewer complications. Also, the system comprises three key modules, including the sensor and executive module, the transcutaneous energy transmission module, and the wireless communication module as shown in Figure 1.

Outline of AAS system.

The sensor and executive module comprises a sensor unit and a novel executive mechanism. The sensor unit placed closely to the anal canal acts in the role of the natural nerve system, which has always monitored the pressure signal of the anal canal and transmitted the signal to the inside controller. The novel executive mechanism which simulates the basic function of human nature anal sphincter is used to prevent unwanted leakage of feces.

Driving energy for the implanted component is provided by transcutaneous energy transmission module based on electromagnetic coupling. Two copper coils separated by the skin are facing each other closely to transmit energy in the maximal degree. Then some regularization must be processed on the initial energy to satisfy the power standard for the whole implanted component.

The communication between the implanted and outside components is realized through wireless communication module. It ensures that the implanted component does not have any wire link with the outside component. Date and command is transmitted through the module to ensure that the system operates normally. This method is the only way to eliminate possible infection caused by wire linking and surgery to replace the battery, which has a great significance for the implanted AAS system.

Figure 2 shows the prototype of the AAS system which integrates the above three modules together. An alarm and display unit is designed to inform patients that the urge to defecate has been detected. As is shown in Figure 2, the alarm and display unit is integrated with the outside wireless communication module in a PCB board. Based on the division of the two copper coils, the left side is regarded as the outside component and the right side is the implanted component. A 12 V experimental rechargeable battery is used to power the entire outside component while the driving energy for the inside component is provided by the transcutaneous energy transmission module.

Prototype of the AAS system.

To simulate a natural defecation process, all the constituent parts of the AAS system must coordinate closely with each other. The mechanism of defecation control can be described by Figure 3. When the urge to defecate is detected, the patient starts the defecation process by just pressing the “start” button. And when the defecation is finished, patients only need to press the “stop” button to shut off the executive mechanism. The whole defecation process is no different from that of people unaffected by fecal incontinence.

Mechanism of defecation control.

The novel executive mechanism

Inspired by the sandwich structure AAS (10-11-12), we designed a hinge structure executive mechanism which is actuated by a DC push-draw electromagnet as shown in Figure 4. The hinge structure is formed by two thin sheets covered by a silicon rubber diaphragm. Composition of the two thin sheets used in this study is medical titanium nickel alloy. The circumferential shape block at the end of the hinge structure is used to replace the anal sphincter muscles and to control the opening and closing of the anal canal. To prevent the probable fistula, the circumferential shape block has a small radian to disperse the compression force around the anal canal. It is made of silicone rubber with softness and high tenacity to reduce compression injury of the anal canal. As the actuator, a DC push-draw electromagnet is associated with the upper thin sheet through a slide block.

The novel executive mechanism.

The novel executive mechanism has a simple structure and its operation is more convenient than the traditional design. When the DC push-draw electromagnet is not electrified, its spring is in a state of compression. The two alloy boards of the hinge structure will be put together under the action of the spring force to clamp the anal canal. When the patient needs to defecate, simply by electrifying the electromagnet, the anal canal can be opened to pass feces. To prevent the mechanism from corroding, a sealing material made of a silicone rubber diaphragm should cover the whole mechanism.

Establishment of prediction model for rectal sensation

Rectal sensation predication model

Related medical research indicates that high-amplitude propagated contractions (HAPC) signals in the biomedical signal can be used to indicate the urge to defecate (13). However, the collected HAPC signal always contains other contractive waves, so the time-frequency domain feature is more conducive for pattern recognition of the rectal signal. Wavelet packet decomposition is an effective time-frequency domain analytical tool. In this paper, an optimal wavelet packet basis based on the DB index is adopted for feature vector extraction. Feature vectors are then inputted to the SVM to predict the urge to defecate. The HAPC signal detected in the rectal signal means occurrence of the urge to defecate. The predication model is described in Figure 5.

Predication model for rectal sensation function rebuilding.

Feature extraction based on wavelet packet decomposition

Many applications of wavelet packet decomposition have been reported in biomedical signal analysis (14-15-16). Wavelet packet decomposition can obtain detailed bandwidth distribution information because it decomposes both the low-frequency and high-frequency parts of rectal signal. In view of the typical nonlinearity and nonstationary characteristics of the rectal signal, the combination of the time and frequency domains can yield more information. Hence, the wavelet packet coefficients mean (time domain information) and the relative energy (frequency domain frequency) in every subspace are employed as the signal features.

Assume that the original rectal signal x(n) is a finite energy signal and J series wavelet packet decomposition. W00 is operated on the signal. Figure 6 shows the tree structure of wavelet packet decomposition. is the original signal space and Wjp(p=0, 1,⋯2j−1) denotes the pth subspace occurring at the jth decomposition level. The wavelet packet coefficients of the subspace Wjp are described by djp={djp(k),(k) = 1, 2,⋯,Kjp} The computation method of wavelet packet coefficients mean and the relative energy in every subspace is presented as described below.

The tree structure of wavelet packet decomposition.

The wavelet packet decomposition coefficient mean MEAj, p in every subspace is adopted as the original coefficients mean feature set M.

M E A j , p = 1 n j , p ∑ k d j p ( k ) ,      [1]

where nj, p is the coefficient length of subspace Wjp. So the original wavelet packet coefficients mean feature vector set is

M = { M E A j , 0 , … M E A j , 1 , … M E A j , p } ,   P = 2 j − 1.      [2]

The relative energy Pj, p in every subspace is adopted as the original relative energy feature set N.

The energy of the pth subspace occurring at the jth decomposition level is

E j , p = ∑ k ( d j p ( k ) ) 2 .      [3]

The total energy of the jth decomposition level is

T E j = ∑ p E j , p .      [4]

The relative energy Pj, p in every subspace is

P j , p = E j , p / T E j .      [5]

The original relative energy feature set is

N = { P j , 0 , P j , 1 , … , P j , p } ,   P = 2 j − 1.      [6]

Hence, we obtain the original feature vector set is X = {M, N}.

Optimal wavelet packet basis based on the Davies-Bouldin (DB) index

It is worth noticing that wavelet packet decomposition provides numerous effective signal decomposition forms. As shown in Figure 6, an effective signal decomposed form consists of W10∪W11,W20∪W21∪W11,W20∪W21∪W22∪W23, etc. Assume that G={G1i=1, 2, 3,…} is the set of all the effective decomposed forms. So G is the wavelet packet bases library and Gi is one of the wavelet packet bases.

Each wavelet packet basis has its unique property that reflects a certain feature of signal. Inadequate or excessive numbers of wavelet packet bases increase the miscellaneous degree of the algorithm. The aim of the optimal wavelet packet decomposition is to select an optimal wavelet packet basis for classification problems from the library of wavelet packet bases. The Davies-Bouldin (DB) index has been proved effective when used to evaluate the classification ability of feature space (17-18-19). Hence, the DB index is adopted in this paper to measure the classification ability and to select the optimal wavelet packet basis.

The DB index is based on the scatter matrices of the data. The DB index is obtained through determining the worst case of separation for each cluster and averaging these values as follows (20):

D B = 1 k ∑ i = 1 k max i ≠ j   R i j ,      [7]

where K is the number of clusters to distinguish and cluster-to-cluster similarity is:

R i j = ( D i i + D j j ) D i j ,      [8]

where Dii and Dj j denotes the dispersions of cluster i and cluster j respectively, Dij is the distance between the mean value of cluster i and cluster j. The value can be obtained through the following formula:

D i i = [ 1 N i ∑ n = 1 N i ∥ y n = m i ∥ 2 ] 1 / 2      [9]

and Dij=‖mi−mj‖,     [10]

where Ni is the number of members in cluster i, yn is the nth sample vector of cluster i, and mi is the mean vector of cluster i. It is obviously that a smaller DB index value indicates a better feature classification ability (20).

In the following, the optimal wavelet packet basis search algorithm based on DB index is given. Suppose set B = (Bjp|j=0,1,…;p=0,1,…,2j−1) and set A = Φ where Bjp is the set of basis vectors belonging to the subspace Wjp.

Step 1: Find the basis vector Bki in set B, where Bki is corresponding to the subspace which has the least DB index value. Then add Bki to set A and remove Bki from B.

Step 2: ∀Bxy∈B, if Bxy is corresponding to the child node space (direct or indirect) or parent node space of Bki, then remove Bxy from set B.

Step 3: If B = Φ, stop the algorithm, else return to Step 2 and continue.

The subspace corresponding to the selected wavelet packet basis must cover the horizontal range completely. Overlap of any subspace in the horizontal direction should be avoided. The final set A is exactly the optimal wavelet packet basis based on the DB index.

Support machine vector (SVM)

Assume that the linearly separable sample set is (xi, yi), i = 1,2,..., n, x ∈ Rd, y ∈ {+1, –1} is class label. The hyperplane in D dimension is described by w · x + b = 0, where w is the weigh vector and b is the offset. For nonlinear problems, it is necessary to map input space to a high-dimensional space by nonlinear transformation x→∅ (x). Then nonlinear classification in low-dimensional space can be replaced by the linear classification problem in the high-dimensional space.

Seeking the best hyperplane can be described as minimization of formula [11]:

F = 1 2 ( w ⋅ w ) + C ( ∑ i = 1 n ξ i ) ,      [11] s . t . y i [ w   ⋅   ∅ ( x i ) + b ] − 1 + ξ i ≥ 0 ,   i = 1 ,   2 , ⋯ ,   n ,

where c is penalty factor and ξi is slack variable. A Lagrange function is introduced to solve the convex secondary optimization problem of formula [11]:

L ( w ,   b ,   ξ ,   α , η ) = 1 2 ‖ w ‖ 2 + c ∑ i = 1 n ξ i − ∑ i = 1 n α i ( y i ( w T ∅ ( X i ) + b ) − 1 + ξ i ) − ∑ i = 1 n η i ξ i ,      [12]

where αi and ηi are Lagrange multipliers. With the partial differential of master variable in formula [14] and formula [13] and formula [14], we got the formula:

W ( α ) = ∑ i = 1 N α i − 1 2 ∑ i = 1 n ∑ j = 1 n α i α j y i y j ∅ T ( X i ) ∅ ( X j ) .      [13]

Kernel function K(xi, xj)=∅T(Xi)∅(Xj) is used to simplify formula [13]. Any functions that satisfy Mercer’s condition could be used as kernel function, while different kernel function would have different classification prediction accuracy. Radial basis function (RBF) kernel function is used in this paper.

K ( x i , x j ) = exp ( − ‖ X i − X j ‖ 2 2 σ 2 ) ,      [14]

where K is the kernel function and σ is kernel function width. Hence, the optimization decision function takes the form:

f ( x ) = sgn ( ∑ i = 1 n α i y i K ( x , x i ) + b ) .      [15]

Experiment analysis

Clamping force measurement experiment

Selecting a suitable clamping force for the executive mechanism is a serious problem urgently needed to be solved. From a theoretical point of view, a large enough force can guarantee unnecessary feces loss. However, long-term oversize force imposed on the rectum easily causes compression injury. Related medical research indicates that the suitable clamping force is 80 mmHg to 120 mmHg. With the aim of reducing harm to human issues as far as possible, a slightly smaller force 10 kPa (about 75 mmHg) is used for the clamping force in this paper. Then its sealing effect is evaluated in the following way.

To evaluate the performance characteristics of the executive mechanism, we established a measurement system which comprises the executive mechanism, a dynamometer, and an adjustable DC power supply. Figure 7 shows the characteristic curve of electromagnet Force-Displacement (F − x) and spring Force-Deformation (F − ∆x). The black line indicates the relationship of spring F − ∆x. The remaining three curves indicate the characteristic curve of electromagnetic F − x respectively under the different working voltage.

Characteristic curve of the electromagnetic and spring forces.

The spring force is proportional to its deformation within the spring elastic deformation range, which manifests as a linear straight line in Figure 7. As for the F − x characteristic, a higher working voltage generates a larger electromagnetic force. When the working voltage is maintained, the electromagnetic force increases as the moveable core approaches the stationary core and the two limit values obviously differ. When the working displacement x is between 4~10 mm, the characteristic curve F − x is approximately horizontal, which means a constant force.

The clamping force imposed on the rectum is created by the spring compressive deformation, so an increase in intestinal contents results in extra compressive deformation, which in turn generates an extra force to keep the closing state. When the electromagnet is working at 12 V voltage with maximum current 0.15 A, the electromagnet temperature rises about 30°C within a continuous 30-minute period, which causes no harm to human tissues. Repeated tests indicate that the sealing effect is good enough to maintain the pressure and prevent feces leakage.

Experiment for rebuilding rectal sensation

The biological parameters telemetry capsule made by Shanghai Jiaotong University was used for rectal pressure signal acquisition of 10 healthy volunteers, including six men and four women. One hundred groups of data with 10 times per person were acquired. Four wrong groups were abandoned, while the remaining 96 groups of data were filtered and de-noised. Then 62 groups of data were selected as training samples and 34 groups of data as test samples. The samples with HAPC signal were regarded as Class 1, which indicates the urge to defect. The samples without HAPC signal were regarded as Class 2, which means no need to prepare for defecation.

The Db3 wavelet with the best regularity properties was chosen as the mother wavelet to decompose the rectal pressure signal. Three-layer wavelet packet decomposition was operated on the rectal pressure signal and 15 different subspaces Wjp were obtained. Then the coefficient mean MEAj, p and relative energy Pj, p in each subspace was obtainable according to Formula [1] and Formula [5] respectively, as shown in Figures 8 and 9.

We adopted three kinds of class 1 signal and three kinds of class 2 signal to extract feature vector to predict the urge to defecate. The feature vector of the Class 1 signal is shown in Figure 8a and Figure 9a; the feature vector of the Class 2 signal is shown in Figures 8b and 9b. As the DB index algorithm is based on the entire wavelet packet decomposition tree, 14 MEAj, p and 14 Pj, p should be calculated. So M is a 14 dimension feature vector of the subspace coefficient mean and N is a 14 dimension feature vector of the subspace relative energy. As is shown in Figures 8 and 9, the subspace number 1, 2, 3,..., 14 denotes the subspace W10,W11,W20,…,W37 respectively. So the original feature vector set X = {M, N} can be obtained, where X is a 28 dimension vector.

Coefficients mean of each subspace. a) Three kinds of Class 1 signal. b) Three kinds of Class 2 signal.

Relative energy of each subspace. a) Three kinds of Class 1 signal. b) Three kinds of Class 2 signal.

The DB index of each subspace in each layer is calculated according to Eq. [7], and then selection of optimal wavelet packet basis is extracted with the optimal wavelet packet basis search algorithm based on the DB index (see above “Optimal Wavelet Packet Basis based on Davies-Bouldin index”). One optimal wavelet packet basis selection process is shown in Figure 10. The value in each subspace is the DB index value of the subspace. The colored subspace is the optimal wavelet packet basis based on DB index, that is W20, W32, W33, W34, W35 and W23. Based on the subspaces corresponding to the selected optimal wavelet packet basis, the MEAj, p and Pj, p which are not belong to the subspaces from feature vector M and N are removed. The goal to simplify the original feature vector set X = {M, N} is thus achieved.

The DB index in each subspace.

In the end, the simplified feature vector X′ = {M′, N′} of 62 groups of training data is inputted into the SVM for model training. Then the 34 groups of test data are inputted to the SVM to testify the effectiveness of the proposed algorithm. The RBF kernel is used in this paper for the SVM. When the HAPC signal is detected in any group of test data, it indicates occurrence of the urge to defecate. The prediction accuracy of the SVM with 34 groups of test sample is shown in Table I. The test prediction accuracy is 97.06% and the operating time is 2.13 s.

PREDICTION ACCURACY OF SUPPORT VIRTUAL MACHINE

Predication model Kernel function Test sample number Correct predication Predication accuracy
SVM = support virtual machine; RBF = radial basis function.
SVM RBF 34 33 97.06%

According to the subspaces corresponding to the optimal wavelet packet basis based on THE DB index, the coefficients mean and relative energy in each subspace was selected as a feature vector to predict the urge to defecate. The predication accuracy can reach up to 97.06%. The experiment results show that the selected feature vectors can indicate the major feature to predict the urge to defecate.

In vitro experimental system

All the modules were integrated to assemble the prototype of the AAS system. The prototype and a human model were combined to form the platform of an in vitro experimental system. Considering that the diameter size of the human anal canal is similar to the small intestine of the pig, the fresh small intestine of a pig is applied to simulate the anal canal which is directly in contact with the executive mechanism. A plastic pipe is applied to simulate the remaining part of the human rectum. The fresh small intestine is connected to the terminal of the plastic pipe and the executive mechanism clamps the terminal of the fresh small intestine. As is shown in Figure 11, the outside component is placed in front of the human model. With a display and alarm unit, the patient can notice the alarm signal when the urge to defecate is detected.

The prototype of the artificial anal sphincter system.

Firstly, the transcutaneous energy transmission module is activated to power the entire system and make sure the wireless communication module is operating normally. In the original state, the executive mechanism is in a closed state to clamp the terminal of the small intestine, as shown in Figure 12a. Then semifluid from the over shedding of the plastic pipe is injected until the fresh small intestine is full with semifluid. At the same time, the executive and sensor module monitors the pressure signal of the small intestine. When the urge to defecate is detected, the executive mechanism opens its hinge structure to pass feces, as shown in Figure 12b.

The experiment on rectal sensation function rebuilding when injecting the semifluid into the in vitro system indicates that the novel executive mechanism can simulate the basic function of the natural anal sphincter, and the proposed algorithm is quite effective for rebuilding the patient’s rectal sensation. a) Closed state to clamp the intestine. b) Open state for passing feces.

CONCLUSIONS

To treat human fecal incontinence, a novel artificial anal sphincter (AAS) system based on rebuilding rectal sensation function is proposed in this paper. A novel executive mechanism with a sandwich structure is designed to simulate the basic function of the natural human anal sphincter. To solve the rectal sensation loss problem caused by fecal incontinence, a method for rebuilding the rectal sensation function is proposed. By extracting the rectal pressure feature with the OWPB based on the DB index, the coefficients mean and relative energy in each subspace after wavelet packet decomposition are taken as feature vectors, and the trained SVM is used to predict the urge to defecate. Experimental results indicate that the novel executive mechanism can simulate the basic function of the nature anal sphincter, and the proposed method is quite effective in rebuilding the patient’s rectal sensation.

Disclosures

Financial Support: This work was supported by National Natural Science Foundation of China (No. 31100708, No. 31370998) and Natural Science Foundation of Shanghai (No.11ZR1412400).
Conflict of Interest: None.
References
  • 1. Bartlett L.,Nowak M.,Ho Y-H. Impact of fecal incontinence on quality of life. World J Gastroenterol 2009; 15: 3276-3282 Google Scholar
  • 2. Vonthein R.,Heimerl T.,Schwandner T.,Ziegler A. Electrical stimulation and biofeedback for the treatment of fecal incontinence: a systematic review. Int J Colorectal Dis 2013; 28: 1567-1577 Google Scholar
  • 3. Leite FR.,Lima MJ.,Lacerda-Filho A. L. FR. Early functional results of biofeedback and its impact on quality of life of patients with anal incontinence. Arq Gastroenterol 2013; 50: 163-169 Google Scholar
  • 4. Devroede G.,Giese C.,Wexner SD. Quality of life is markedly improved in patients with fecal incontinence after sacral nerve stimulation. Female Pelvic Med Reconstr Surg 2012; 18: 103-112 Google Scholar
  • 5. Skull A.,Hull TL. Sacral nerve stimulation for fecal incontinence. Expert Rev Med Devices 2012; 9: 477-482 Google Scholar
  • 6. Khaikin M.,Wexner SD. Treatment strategies in obstructed defecation and fecal incontinence. World J Gastroenterol 2006; 12: 3168-3173 Google Scholar
  • 7. Chen TA. Loop ileostomy or loop colostomy: which one is better for fecal diversion? Int J Colorectal Dis 2012; 27: 131-132 Google Scholar
  • 8. Lehur PA.,Roig JV.,Duinslaeger M. L. P.A. Artificial anal sphincter: prospective clinical and manometric evaluation. Dis Colon Rectum 2000; 43: 1100-1106 Google Scholar
  • 9. Malouf AJ.,Vaizey CJ.,Kamm MA.,Nicholls RJ. Reassessing artificial bowel sphincters. Lancet 2000; 355: 2219-2220 Google Scholar
  • 10. Luo Y.,Okuyama T.,Takagi T. Thermal control of shape memory alloy artificial anal sphincters for complete implantation. Smart Mater Struct 2005; 14: 29-35 Google Scholar
  • 11. Luo Y.,Takagi T.,Okuyama T. Functional evaluation of an artificial anal sphincter using shape memory alloys. ASAIO J 2004; 50: 338-343 Google Scholar
  • 12. Luo Y.,Higa M.,Amae S. The possibility of muscle tissue reconstruction using shape memory alloys. Organogenesis 2005; 2: 2-5 Google Scholar
  • 13. Liem O.,Burgers RE.,Connor FL. Solid-state vs water-perfused catheters to measure colonic high-amplitude propagating contractions. Neurogastroenterol Motil 2012; 24: 345-e167 Google Scholar
  • 14. Safara F.,Doraisamy S.,Azman A. Wavelet packet entropy for heart murmurs classification. Adv Bioinf 2012; 2012: 327269- Google Scholar
  • 15. Strauss D.,Jung J.,Rieder A.,Manoli Y. Classification of endocardial electrograms using adapted wavelet packets and neural networks. Ann Biomed Eng 2001; 29: 483-492 Google Scholar
  • 16. Zalay OC.,Kang EE.,Cotic M.,Carlen PL.,Bardakjian BL. A wavelet packet-based algorithm for the extraction of neural rhythms. Ann Biomed Eng 2009; 37: 595-613 Google Scholar
  • 17. Godiyal AK.,Sharma R.,Joshi D.,Bhatia D. Feasibility of facial EMG in gender classification during speech production. J Med Eng Technol 2013; 37: 86-90 Google Scholar
  • 18. Khadivi Heris H.,Seyed Aghazadeh B.,Nikkhah-Bahrami M. Optimal feature selection for the assessment of vocal fold disorders. Comput Biol Med 2009; 39: 860-868 Google Scholar
  • 19. Wang G.,Ren D. Classification of surface electromyographic signals by means of multifractal singularity spectrum. Med Biol Eng Comput 2013; 51: 277-284 Google Scholar
  • 20. Wang G.,Wang Z.,Chen W.,Zhuang J. Classification of surface EMG signals using optimal wavelet packet method based on Davies-Bouldin criterion. Med Biol Eng Comput 2006; 44: 865-872 Google Scholar

Authors

  • Zan, Peng [PubMed] [Google Scholar]
  • Liu, Jinding [PubMed] [Google Scholar]
  • Jiang, Enyu [PubMed] [Google Scholar]
  • Wang, Hua [PubMed] [Google Scholar]

Affiliations

  • School of Mechatronics Engineering and Automation of Shanghai University, Shanghai Key Laboratory of Power Station Automation Technology, Shanghai - PR China

Article usage statistics

The blue line displays unique views in the time frame indicated.
The yellow line displays unique downloads.
Views and downloads are counted only once per session.

No supplementary material is available for this article.