A Novel MOGA-SVM Multinomial Classification for Organ Inflammation Detection

Kwok Tai Chui; Miltiadis D. Lytras

doi:10.3390/app9112284

A Novel MOGA-SVM Multinomial Classification for Organ Inflammation Detection

Chui, Kwok Tai;Lytras, Miltiadis D. 2019-06-03 00:00:00 applied sciences Article A Novel MOGA-SVM Multinomial Classiﬁcation for Organ Inﬂammation Detection 1 , 2 , 3 Kwok Tai Chui * and Miltiadis D. Lytras Department of Electronic Engineering, City University of Hong Kong, Hong Kong SAR, China School of Business & Economics, Deree College—The American College of Greece, 6 Gravias Street GR-153 42, Aghia Paraskevi, 15342 Athens, Greece; [email protected] Eat College of Engineering, Eat University, Jeddah P.O. Box 34689, Saudi Arabia * Correspondence: [email protected] Received: 6 April 2019; Accepted: 29 May 2019; Published: 3 June 2019 Featured Application: In this paper, a novel multi-objective genetic algorithm based support vector machine (MOGA-SVM) has been proposed. A customized similarity kernel has been optimally designed for the multinomial classiﬁcation of the inﬂammations of appendix, pancreas, and duodenum. Practically, this methodology can be applied to other classiﬁcation problems as the concept of the methodology is to customize the kernel to speciﬁc application. In order to achieve a better performance using kernel based algorithm, it is highly recommended to use customize kernel instead of traditional kernels. Abstract: Wrist pulse signal (WPS) contains crucial information of humans’ health condition. It can serve as an alternative method for diagnosing of organ inﬂammation instead of traditional clinical measurement. In this paper, a novel multi-objective genetic algorithm based support vector machine (MOGA-SVM) has been proposed for the multinomial classiﬁcation of the inﬂammations of appendix, pancreas, and duodenum. A customized similarity kernel (K ) has been optimally designed. The CS performance of multinomial classiﬁcation using K is compared with ﬁve types of kernels, linear, CS radial basis function (RBF), polynomial and sigmoid kernel, as well as mixtures of polynomial and RBF, to verify the eectiveness of K . The sensitivity, speciﬁcity and accuracy (Acc) of the proposed CS method are 92%, 91.2%, and 91.6% respectively. The results have demonstrated that K improves CS the accuracy of classiﬁcation from 8.9% to 59.6%. When compared to related work, the proposed method increases the performance by more than 10%. It is believed that WPS can serve as alternative measures to diagnose organ inﬂammations. Keywords: bioinformations; genetic algorithm; multiobjective optimization; organ inﬂammation; support vector machine; wrist pulse signal 1. Introduction Health is crucial element in today’s life. Researchers have devoted vast eorts in proposing new policies, algorithms, systems, and architectures for healthcare. According to the World Health Organization (WHO), in 2013, the global requirement and the actual number of health workforce were 60.4 million and 43 million, respectively [1]. These ﬁgures will be increased to 81.8 million and 67.3 million, respectively, by 2030. Hence, it is believed that the shortage of medical personnel is unsolved and remained serious in the coming decade. Automatic decision making via machine learning is believed to be the only way out to solve the shortage of medical personnel [2,3]. Medical workers may argue that the automatic system has a conﬂict of interest with them; nevertheless, it is not the truth. First, the current workload of medical workers (ratio of workers to patients) is heavy Appl. Sci. 2019, 9, 2284; doi:10.3390/app9112284 www.mdpi.com/journal/applsci Appl. Sci. 2019, 9, 2284 2 of 14 and will become normal. Second, an automatic system focuses on routine works, so that medical workers can devote more time to professional consultation and surgery activities. Third, the increase in quality of medical services will lead to higher acceptance and satisfaction by the public. Thus, medical workers will earn a higher social status and better job satisfaction. Many diseases and abnormal human conditions can be examined by digital imaging diagnostic, like X-ray, Computed Tomography (CT) scan, Magnetic Resonance Imaging (MRI), Ultrasonography, Electrocardiogram, and Biopsy. In this paper, Wrist pulse signal (WPS) of human is considered which provides key information regarding health conditions. In the literature, WPS can be utilized for various applications, for instance, pre-meal and post-meal classiﬁcation [4], physical exercise [5], diabetes classiﬁcation [6], hypertension association [7,8], lung cancer recognition [9], and inﬂammation classiﬁcation [10,11]. Various signal processing techniques on WPS can be found in [12–14], for instance, dynamic time warping, wavelet analysis, periodic decomposition, principal component analysis, and linear discriminant analysis. In this paper, four common types of organ inﬂammation are considered, namely, appendicitis, acute appendicitis, duodenitis, and pancreatitis. According to the WHO, the annual deaths hat are attributable to appendicitis, duodentitis and pancreatitis in 2000, 2005, 2010, and 2015 are shown in Table 1 [15]. It is noted that acute appendicitis is embedded into Global Health Expand (GHE) code 1240. From Table 1, the number of deaths in each category is increasing by an increment of29%, 24%, and 60% for appendicitis, duodenitis, and pancreatitis, respectively, from 2000 to 2015. Among three types of organ inﬂammations, pancreatitis is the leading cause, which is followed by duodenitis and appendicitis. To conclude, the issues of deaths in these organ inﬂammations remain unsolved. Table 1. Number of deaths due to Appendicitis, Pancreatitis, and Duodenitis in 2000, 2005, 2010, and 2015. Number of Deaths (Annual) GHE Cause GHE Code 2000 2005 2010 2015 1240 Appendicitis 34,800 39,400 43,300 45,000 1241 Duodenitis 37,900 40,400 43,800 47,000 1248 Pancreatitis 64,400 77,800 93,900 103,500 There have been more than million of suerers and thus it is necessary to have a reliable and accurate method for the diagnosis of organ inﬂammations. Based on literature ﬁnding, there are a few publications working on binary classiﬁcation of healthy, appendicitis, acute appendicitis, duodenitis, and pancreatitis suerers [10,11]. In [10], the features extraction process, an auto-regression (AR) based model was proposed. Two features, the standard deviation and mean of the prediction error from AR model, were chosen to represent the information of the WPS, and for further analysis. With regard to the classiﬁcation, the support vector machine (SVM) with linear kernel was adopted for the binary classiﬁcation, which yields an accuracy of 77.8–91.2%. For further improvement, a radial basis kernel (RBF) has been utilized to replace the linear kernel [11]. The idea is that most of the classiﬁcation problems are not linearly separable. The enhanced method achieved an accuracy of 88.6–98.4%. Nevertheless, as a pragmatic application, it is deemed to be formulated as classifying instances into one of the more than two classes, and multinomial classiﬁcation is desired. A novel multi-objective genetic algorithm (MOGA) based SVM, abbreviated as MOGA-SVM, has been proposed for the multinomial classiﬁcation of the organ inﬂammations of appendicitis, acute appendicitis, duodenitis, and pancreatitis. MOGA is a heuristic approach that has been widely adopted to obtain tradeo solutions between two or more conﬂicting objectives [16–18]. SVM receives a lot of attention as a supervised learning algorithm for classiﬁcation problems [19–21]. In this paper, a customized similarity kernel (K ) has been optimally designed for speciﬁc application, the classiﬁcation of appendicitis, CS acute appendicitis, duodenitis, and pancreatitis. It is worth mentioning that traditional kernels, like linear, RBF, quadratic, and polynomial kernels are not designed for any particular application. It is Appl. Sci. 2019, 9, 2284 3 of 14 recommended that the customized kernel should be utilized for organ inﬂammations classiﬁcation instead of traditional kernels in order to improve the classiﬁcation accuracy. This paper is organized, as follows. Section 2 provides the background of organ inﬂammations and an overview of MOGA-SVM. The methodology of the proposed algorithm is explained in Section 3. Performance evaluation and comparison are given in Section 4. Finally, a conclusion is made in Section 5. 2. Dataset and Overview of MOGA-SVM The background symptoms of each organ inﬂammation, appendicitis, acute appendicitis, duodenitis, and pancreatitis will ﬁrstly describe. Only a summary is provided in each topic. Readers who are interested in the details of the inﬂammations are suggested to refer to appendicitis [22], acute appendicitis [23], duodenitis [24], and pancreatitis [25]. Subsequently, the overview of the MOGA-SVM is brieﬂy discussed, in which the details will be explained in the next section. 2.1. Background of Organ Inﬂammation 2.1.1. Appendicitis Appendicitis is an inﬂammation of the appendix. It is not uncommon abdominal emergency at any age. The causes are due to the blocking of appendix by stool, cancer, or foreign body, or from infection. Typical symptoms are abdominal pain, loss of appetite, diarrhea, and inability to pass gas. The clinical approaches for diagnosis include abdominal exam, urine test, rectal exam, blood test, CT scans, and ultrasound. The lifetime risk of suering from appendicitis is about 7%, with dierent severity levels [22]. The occurrence of this inﬂammation is approximately 11 persons per 10,000 populations in each year. 2.1.2. Acute Appendicitis Acute appendicitis is more severe than appendicitis, which has annual incidence of 90 to 140 per 10,000 populations [23]. Although this inﬂammation has been documented for more than 500 years, its etiology is not well known. It is usually results from injury of its mucosa and spread from that injury via its wall. The symptoms and examinations of acute appendicitis are similar to that in appendicitis. 2.1.3. Duodenitis Duodenitis is inﬂammation of the duodenum. The known causes include helicobacter pylori infection, bacterial infection, Nonsteroidal anti-inﬂammatory drug, viral infection, coeliac disease, and idiopathic [24]. Abdominal pain, nausea, vomiting, and discomfort in stomach are the four known symptoms. The most common examination is an Oesophago-Gastro-Duodenoscopy. For the global annual years of healthy life lost, the estimation is about 58 persons per 100,000. 2.1.4. Pancreatitis Pancreatitis is inﬂammation of the pancreas. It is more important than the aforementioned three organ inﬂammations, because it often characterized by irreversible change, permanent loss of function [25]. The clinical features of pancreatitis include ﬁbrosis, chronic and recurrent inﬂammation, duct distortion, atrophy, and the risk of pancreatic cancer. The estimated incidence of pancreatitis is 42 persons per 100,000 population [26]. 2.2. Overview of MOGA-SVM Figure 1 shows the ﬂow chart of MOGA-SVM for organ inﬂammations classiﬁcation. The typical waveform of the wrist pulse signal is shown in Figure 2, which is characterized by a percussion wave, tidal wave, dicrotic wave, peak systolic velocity, reverse velocity, peak diastolic velocity, and end diastolic velocity. Appl. Sci. 2019, 9, x 5 of 15 Appl. Sci. 2019, 9, 2284 4 of 14 Appl. Sci. 2019, 9, x 5 of 15 Figure 1. Overview of multi-objective genetic algorithm based support vector machine (MOGA-SVM) Figure 1. Overview of multi-objective genetic algorithm based support vector machine (MOGA-SVM) Figure 1. Overview of multi-objective genetic algorithm based support vector machine (MOGA-SVM) for organ inflammations classification. for organ inflammations classification. for organ inﬂammations classiﬁcation. Figure 2. Typical waveform of wrist pulse signal. The similarity coefficients of every pair of WPS are computed, which form the customized Figure 2. Typical waveform of wrist pulse signal. Figure 2. Typical waveform of wrist pulse signal. similarity kernel. After MOGA, the optimal kernel KCS is designed. A classifier for organ inflammations classification is constructed. Section 3discusses the details (Figure 3 is drawn to The training of organ inﬂammations classiﬁer, the datasets contain WPS of healthy, appendicitis, The similarity coefficients of every pair of WPS are computed, which form the customized summarize the key steps of the MOGA-SVM). In this paper, the 10-fold cross-validation is adopted acute appendicitis, duodenitis, and pancreatitis candidates were considered [10]. Each record of WPS similarity kernel. After MOGA, the optimal kernel KCS is designed. A classifier for organ to evaluate the classifier, as it is a practical order in literature [27,28]. is carried out DC drift elimination and low-pass ﬁlter following the approach, as in [11]. Afterwards, inflammations classification is constructed. Section 3discusses the details (Figure 3 is drawn to the local maxima and minima points of the WPS are located. The detail is not being discussed in this summarize the key steps of the MOGA-SVM). In this paper, the 10-fold cross-validation is adopted paper, as the authors would like to mainly focus on the proposed MOGA-SVM. to evaluate the classifier, as it is a practical order in literature [27,28]. Appl. Sci. 2019, 9, 2284 5 of 14 The similarity coecients of every pair of WPS are computed, which form the customized similarity kernel. After MOGA, the optimal kernel K is designed. A classiﬁer for organ inﬂammations CS Appl. Sci. 2019, 9, x 9 of 15 classiﬁcation is constructed. Section 3 discusses the details (Figure 3 is drawn to summarize the key steps of the MOGA-SVM). In this paper, the 10-fold cross-validation is adopted to evaluate the classiﬁer, generations or if the output reaches the pareto front. To facilitate readers’ understanding, Algorithm as it is a practical order in literature [27,28]. 1 and Algorithm 2 are presented as the pseudo code of MOGA-SVM. Figure 3. Optimal design of KCS and classifiers using MOGA-SVM. Figure 3. Optimal design of K and classiﬁers using MOGA-SVM. CS A pseudo code for the MOGA-SVM is given for better illustration and understanding. When it comes to practical application, the WPS of the candidate is measured and it served as the Algorithm 1 Segmentati on ( X ) input of the trained organ inﬂammations classiﬁer. The outputs maintain ﬁve possibilities, healthy, appendicitis, acute appendicitis, duodenitis, and pancreatitis. If the status is one of the four organ Data: Organ inflammations of appendicitis, acute appendicitis, duodenitis and pancreatitis retrieved inﬂammations, from 248 cand aid rat eport es [1will 0], Xbe m sent to a nurse and doctor for further examination and treatment. Output: WPS samples Xi,j 3. Methodology Step 1: dc drift elimination Step 2: Filter Xm using low pass filter Hlow This section is composed of three parts. First, the datasets of healthy, appendicitis, acute Step 3: Locate local maxima and minima points of the Xm; appendicitis, duodenitis, and pancreatitis candidates are illustrated in Section 3.1. Next, the data Step 4: Locate two maxima points with interval of 120 sampling points; preprocessing of the datasets is explained in Section 3.2. At last, Section 3.3 formulates the optimal Xi,j (i=1:4=class label, j=length(Class))←Portion of signal between two maxima points with interval of design of the K . CS 120 sampling points 3.1. Datasets of Organ Inﬂammations Classiﬁer Algorithm 2TrainClass ifier (Classlabel , K , K ) c cc Gratitude is expressed to researchers in [10] for sharing the datasets. The WPSs were measured Data: Classlabel, Kc, Kcc while using Doppler ultrasonic blood analyzer module. In each measurement, three steps were Output: Model followed. Firstly, an approximated position was located where the ﬂuctuation of signal was larger Step 1: generations = 1 than the other positions. Subsequently, a ﬁne tuning of position with slight variation of angle and Step 2: initialization (population) position until the largest signal was observed. Finally, WPS was recorded under the setting of largest Step 3: Evaluate the individuals with the fitness function (F1 and F2) signal amplitude. Step 4: rank the individuals by their fitness values by step 3 Table 2 summarizes the details of the datasets. Assign the class label to each of the category, Step 5: do the Niche count calculation Class 0: healthy, Class 1: appendicitis, Class 2: acute appendicitis, Class 3: duodenitis, and Class 4: while generations <= max_generationdo pancreatitis. The datasets are formed by four age groups, [0, 20), [20, 40), [40, 60) and [60, 100). The Step 6: Select two parents from the population total number of samples is 248 and the corresponding samples in Class 0–4 are 100, 22, 38, 42, and Step 7: Create the offspring using Roulette wheel selection,crossover and mutation 46, respectively. Step 8: Train SVM model for each individual Step 9: Evaluate the offspring with the fitness function (F1and F2) Step 10: rank the individuals by their fitness values by step 3 Step 11: do the Niche count calculation Appl. Sci. 2019, 9, 2284 6 of 14 Table 2. Sample distribution of the datasets. Age Class Name Total [0,20) [20,40) [40,60) [60,100) 0 Healthy 8 26 30 16 100 1 Appendicitis 0 22 0 0 22 2 Acute Appendicitis 20 8 10 0 38 3 Duodenitis 4 26 6 6 42 4 Pancreatitis 16 26 4 0 46 3.2. Data Preprocessing The data preprocessing of the aforementioned samples is following the related work [11]. It includes DC drift elimination, six-order Butterworth low-pass ﬁlter, and the detection of local maxima and minima points. In this analysis, the WPS has a cycle less than 120 samples. The individual sample is formed by the portion between the two largest maxima points. There are 1800, 630, 972, 1386, and 828 samples for healthy, appendicitis, acute appendicitis, duodenitis, and pancreatitis candidates, respectively. For equal division using 10-fold cross validation, two, six, and eight samples have been removed for acute appendicitis, duodenitis, and pancreatitis candidates. Overall, there are 5600 samples. 3.3. Formulation of Optimal K and MOGA-SVM Classiﬁer CS Kernel is essential in SVM classiﬁcation and it has to obey Mercer ’s theorem. That is, the kernel is positive semi-deﬁnite. A common interpretation of kernel is that it captures the correlation between pairs of data. Thus, the proposed K is optimally designed using convolution and cross-correlation. CS The K is formulated as multi-objective optimization problem and is solved by MOGA [29]. CS Let X (n) of length 120 (zero padding for length<120) be the WPS sample. The subscript i refers to i,j the class label from 0 to 4 and that of j refers to the sample number. Therefore, the sets in Class 0 to Class 4 are {X (n), ::: ,X (n)}, {X (n), ::: ,X (n)},{X (n), ::: ,X (n)},{X (n), ::: ,X (n)},and 0,1 0,1800 1,1 1,630 2,1 2,970 3,1 3,1380 {X (n), ::: ,X (n)}, respectively. 4,1 4,820 The convolution between two WPSs X (n) and X (n) is given by a,b c,d N1 a,b C (n) = X (n) X (n) = X (k)X (n k) (1) a,b c,d a,b c,d c,d k=0 where N = 120 is the length of the WPS sample. The cross-correlation between two WPSs X (n) and X (n) can be expressed as a,b c,d N1 X (n)X (n k), k 0 > a,b c,d a,b n=k R (k) = (2) > Njkj1 c,d X (n)X (n k), k < 0 a,b c,d n=0 The customized similarity kernel K is formulated by customized convolution kernel K and CS c customized cross-correlation kernel K . K , and K are deﬁned as cc c cc 2 3 6 X X 7 c,1,1 c,1,N 6 t 7 6 7 6 7 6 . . 7 6 7 . . . K = (3) 6 7 c . 6 . . 7 6 7 6 7 4 5 X X c,N ,N c,N ,1 t t t Appl. Sci. 2019, 9, 2284 7 of 14 2 3 6 X X 7 cc,1,1 cc,1,N 6 t 7 6 7 6 7 6 . . 7 6 7 . . . K = (4) 6 7 cc . 6 . . 7 6 7 6 7 4 5 X X cc,N ,1 cc,N ,N t t where N = 5040 is the 90% of the training samples in Class 0 to Class 4. Here, X refers to the t c,i,j weighting sum of convolution coecients between ith and jth WPS sample. The 1st to 1620th samples come from Class 0. The 1621th to 2187th samples come from Class 1. The 2188th to 3060th samples come from Class 2. The 3061th to 4302th samples come from Class 3. The 4303th to 5040th samples come from Class 4. X and X are computed by c,i,j cc,i,j 2N1 a,b X = w C (m) (5) c,i, j c,m c,d m=1 2N1 a,b X = w R (m) (6) cc,i, j cc,m c,d m=1 where w and w are the weightings of convolution coecients and cross-correlation coecients, c,m cc,m respectively. These weightings are optimally designed while using MOGA. The kernels K and K predominantly aect the maximum margin and the accuracy (Acc) of the c cc organ inﬂammations classiﬁer. From (3) and (4), the maximum margin is directly related to both X ,i,j and X . Based on (5) and (6), an optimal design of both X and X can be obtained by an optimal cc,i,j x,i,j cc,i,j a,b design of the weighting sw and w form = 1, ::: ,2N1, for the given sequences of C (m) and c,m cc,m c,d a,b R (m), respectively. In general, an optimally designed kernel will speed up the convergence of the c,d training algorithm MOGA-SVM. However, varying the combinations of w and w will deduce the c,m cc,m dierent kernels. Searching the optimal weightings require a large computational power. As a result, there is a tradeo between the accuracy and computational power. As it is dicult to ﬁnd the optimal values of w and w that are attributable to the complexity of the objective function, a good trial of c,m cc,m w and w are primarily important, which determines the accuracy. c,m cc,m In this paper, an multi-objective optimization approach, MOGA, is employed to determine the weighting sw and w . However, exhaustive search algorithms may not be the appropriate choices c,m cc,m for searching the solution of the optimization problem. The reason is that it deals with a huge range of combinations. Indeed, heuristic search algorithms eciently and eectively perform the searching of the optimal solutions. In particular, the GA is a robust searching heuristic algorithm that imitates the process of the natural evolution for searching the solution of the optimization problem by the operations in selection, inheritance, crossover and mutation. Aforementioned, to be a proper kernel for SVM classification, the fulfillment of Mercer ’s theorem is essential [30]. The evaluation of eigenvalues helps to determine the positive semi-definite of K and K . c cc K V = D V c c c c (7) K V = D V cc cc cc cc where V and V are non-zero eigenvectors for K and K , respectively. D and D are the c cc c cc c cc corresponding eigenvalues. All of the eigenvalues must be positive in order to ensure K and K are positive semi-deﬁnite. cc Deﬁne K as the sum of K and K . It takes the advantageous from K and K . It is proved below cs c cc c cc that the sum of Mercer ’s kernels is also a Mercer ’s kernel. If K and K are positive semi-deﬁnite, then c cc n T T for any c 2 < , c K c 0 and c K c 0. Hence, c cc T T c K c = c (K + K )c (8) CS c cc Appl. Sci. 2019, 9, 2284 8 of 14 T T T c K c = c K c + c K c 0 (9) CS c cc Therefore, the K is positive semi-deﬁnite. Thus, it is a Mercer ’s kernel. The multi objective CS optimization problem is formulated with two objective functions: Max F = M(, w) (10) Max F = OA = 0.5(S + S ) 2 e p > 0, y = 0, i = 1,::: , N i i i > i=1 2N1 2N1 P P s.t.> (11) > w = 1, w = 1 c,n cc,n n=1 n=1 D 0, D 0,8i c,i cc,i where M(, w) is the margin function of the classiﬁer, is the Lagrange multiplier, S is the sensitivity, i e S is the speciﬁcity, y 2 f1, +1g is the output of the classiﬁer, and D and D are the entries of D p i c,i cc,i c and D , respectively. The margin function is deﬁned as cc N N N X X X M(, w) = y y K (12) i i j i j CS i=1 i=1 j=1 Figure 3 shows the MOGA-SVM for the optimal design of the K . The procedures are as follows: CS (i) The population size and values of objective function are initialized; (ii) The values of objective function of individuals in the population are computed while using the values of objective function deﬁned in (i); (iii) Ranking the individuals according to the values of objective function; (iv) The population convergence is dependent on small group of pareto optimal solutions, but not all optimal solutions are attributable to the nature of the stochastic selection errors, given a limited population size; (v) Niche count is introduced to enhance the population diversity by lengthening the distance between two optimal solutions along the axis of objective functions. The convergence to small group solutions will be avoided; (vi) A new ospring is generated and the values of the objective functions are evaluated; (vii) Ranks assignment and niche count calculation are carried out repeatedly in the new ospring; and, (viii) The algorithm is terminated if it attains the maximum number of generations or if the output reaches the pareto front. To facilitate readers’ understanding, Algorithm 1 and Algorithm 2 are presented as the pseudo code of MOGA-SVM. A pseudo code for the MOGA-SVM is given for better illustration and understanding. Algorithm 1 Segmentation(X Data: Organ inﬂammations of appendicitis, acute appendicitis, duodenitis and pancreatitis retrieved from 248 candidates [10], X Output: WPS samples X i,j Step 1: dc drift elimination Step 2: Filter X using low pass ﬁlter H m low Step 3: Locate local maxima and minima points of the X ; Step 4: Locate two maxima points with interval of 120 sampling points; X (i = 1:4 = class label, j = length(Class)) Portion of signal between two maxima points with interval of 120 i,j sampling points Appl. Sci. 2019, 9, 2284 9 of 14 Algorithm 2 TrainClass i f ier(Classlabel, K , K ) c cc Data: Classlabel, K , K c cc Output: Model Step 1: generations = 1 Step 2: initialization (population) Step 3: Evaluate the individuals with the ﬁtness function (F1 and F2) Step 4: rank the individuals by their ﬁtness values by step 3 Step 5: do the Niche count calculation while generations <= max_generation do Step 6: Select two parents from the population Step 7: Create the ospring using Roulette wheel selection,crossover and mutation Step 8: Train SVM model for each individual Step 9: Evaluate the ospring with the ﬁtness function (F1and F2) Step 10: rank the individuals by their ﬁtness values by step 3 Step 11: do the Niche count calculation Step 12: Decide the new population based on the ospring Step 13: generations = generations + 1 End while Model Pareto solutions 4. Performance Evaluation and Comparison Section 4 is divided into three sub-sections. Firstly, the performance of the proposed K is CS analyzed. Afterwards, it is compared with ﬁve other kernels using the feature extraction approach. Finally, performance comparison between proposed and related work is discussed. 4.1. Performance of Proposed MOGA-SVM Using K CS The performance evaluation of the proposed MOGA-SVM using K adopts 10-fold cross CS validation. Randomly divide 5600WPS samples into 10 equal-sized subsets; with each set containing 560 (10%) samples with Class 0: 180 samples, Class 1: 63 samples, Class 2: 97 samples, Class 3: 138 samples, and Class 4: 82 samples. In each fold of validation, 90% of datasets (nine subsets) from each class serves as training dataset and 10% of the remaining subset serves as the testing datasets. This process completes one-fold of operations. Subsequently, another set is chosen for validation and the remaining nine subsets are used for training. It is noted that this chosen validation set must be dierent from the validation sets that were selected in the previous folds of operations. The process is repeated until all of the 10 subsets have been validation. Applying 10-fold cross validation, the proposed MOGA-SVM using KCS achieves average S , S , e p and Acc of 92%, 91.2%, and 91.6%, respectively. 4.2. Evaluation of Other Kernels Using Feature Extraction Approach In this subsection, feature extraction using convolution coecients and cross-correlation coecients as features is adopted. The following ﬁve kernels, linear, RBF polynomial and sigmoid kernel, and mixtures of polynomial and RBF kernels [31] are applied. They can be expressed by: D E Linear kernel : k (x , x ) = x , x (13) 1 i j i j RBF kernel : k (x , x ) = exp( x x /2) (14) 2 i j i j 2 D E Polynomial kernel : k (x , x ) = x , x + c (15) i j i j D E Sigmoid kernel : k (x , x ) = tanh x , x + c (16) 4 i j i j Mixtures of polynomial and RBF kernels k (x , x ) = k (x , x ) + (1 )k (x , x ) (17) 5 i j 3 i j 4 i j Appl. Sci. 2019, 9, 2284 10 of 14 Three scenarios are considered: (i) Only convolution coecients serve as features (1–199 coecients); (ii) Only cross-correlation coecients serve as features (1–199 coecients); and, (iii) Both convolution and cross-correlation coecients serve as features (1–398 coecients). Table 3 summarizes the performance of kernels K –K in three scenarios. Only the best scenario 1 5 is given. The results reveal that scenario (iii) achieves highest performance, because it takes the advantages from both the convolution and cross-correlation coecients. Compared K with K –K , CS 1 5 the ranking (from highest to lowest) is K > K > K > K > K > K . When compared to scenarios (i), CS 5 3 2 4 1 (ii), and (iii), K improves the Acc by 14.4–58.2%, 12.4–59.6%, and 8.9–53.7%, respectively. CS Table 3. Analysis of traditional kernels in organ inﬂammation classiﬁcations. Performance Kernel Scenario (i) Scenario (ii) Scenario (iii) (S ,S ,Acc)% (S ,S ,Acc)% (S ,S ,Acc)% e p e p e p k (x ,x ) (57.6, 58.2, 57.9) (57.7, 57.1, 57.4) (58.8, 60.4, 59.6) 1 i j k (x ,x ) (76.7, 77.5, 77.1) (76.8, 76.6, 76.7) (77.3, 78.3, 77.8) 2 i j k (x ,x ) (77.6, 78.2, 77.9) (78.3,78.9, 78.6) (78.7, 80.1, 79.4) 3 i j k (x ,x ) (73.8, 74.6, 74.2) (73.2, 73.0, 73.1) (74.8, 75.8, 75.3) 4 i j k (x ,x ) (79.9, 80.3, 80.1) (82.0, 81.0, 81.5) (83.8, 84.4, 84.1) 5 i j 4.3. Comparison between Proposed and Related Work Based on our ﬁnding, the multinomial classiﬁcation of appendicitis, acute appendicitis, duodenitis, and pancreatitis is the ﬁrst of its kind. Previous works [10,11] have considered the problem as binary classiﬁcation. To compare the performance between the proposed and related work [10,11], it is analyzed in two directions. (i) Table 4 gives the raw comparison between the works. (ii) Table 5 gives the matched comparison between the works. The forms of the datasets, application, and cross-validation in [10,11] will be changed into those in this paper. Thus, every work considers 5600 samples for multinomial classiﬁcation and evaluates using 10-fold cross validation. From the raw comparison, it can be seen that the performance, S , S , and Acc of classiﬁcation e p between [10] and [11] are similar for binary classiﬁcation between healthy and appendicitis candidates, and between healthy and duodenitis candidates. For that between healthy and acute appendicitis, and between healthy and pancreatitis, the improvements are 8% and 9%, respectively. If the proposed work is taken into account, it outperforms [10] in the classiﬁcation of all inﬂammations. By averaging the Acc in [11], it is approximately equal to the proposed work. Therefore, it can be interpreted that multinomial classiﬁcation can be achieved without deteriorating the performance in inﬂammations classiﬁcation. A matched comparison environment is setup to compare the performance between algorithms in organ inﬂammations classiﬁcation. Repeated simulation is carried out for [10,11,14,32] while using the identical datasets and 10-fold cross validation. It is concluded that the proposed MOGA-SVM improves the Acc from 6.9% to 13.4%. Appl. Sci. 2019, 9, 2284 11 of 14 Table 4. Raw comparison between proposed and related work [10,11]. Cross Work Method Feature Extraction Dataset (Samples) Class Labels S (%) S (%) Acc (%) e p Validation Class 0: healthy; 81.8 93.3 91.2 Class 1: appendicitis Class 0: healthy; Healthy (100), 76.5 82.4 80.8 Binary Classiﬁcation Class 1: acute appendicitis Mean and standard appendicitis (22), acute using modiﬁed Class 0: healthy; deviation of appendicitis (38), [10] No auto-regressive model 80.0 91.4 88.0 Class 1: duodenitis prediction error duodenitis (42) and and linear kernel pancreatitis (46) SVM Class 0: healthy; 83.3 94.4 90.9 Class 1: pancreatitics Class 0: healthy; 80.4 89.7 87.3 Class 1: All inﬂammations Class 0: healthy; N/A N/A 92.8 Class 1: appendicitis peak systolic velocity; Healthy (100), reverse velocity; peak Class 0: healthy; appendicitis (100), N/A N/A 88.1 diastolic velocity; end Class 1: acute appendicitis Binary Classiﬁcation [11] acute appendicitis (100), diastolic velocity; 10-fold using RBF SVM Class 0: healthy; duodenitis (100) and duration of systole; N/A N/A 88.6 Class 1: duodenitis pancreatitis (100) and duration of diastole Class 0: healthy; N/A N/A 98.4 Class 1: pancreatitics Healthy (1800), Class 0: health; Multinomial Cross-correlation and Appendicitis (630), Class 1: appendicitis; Our work Classiﬁcation using convolution Acute Appendicitis 10-fold Class 2: acute appendicitis; 92.0 91.2 91.6 customized kernel coecients (970), Duodenitis (1380) Class 3: duodenitis; and Pancreatitis (820) Class 4: pancreatitis Appl. Sci. 2019, 9, 2284 12 of 14 Table 5. Matched comparison between proposed and related work [10,11,32]. Cross Work Method Feature Extraction Dataset (Samples) Class Labels S (%) S (%) Acc (%) e p Validation Binary Classiﬁcation Healthy (1800), Class 0: healthy; using modiﬁed Mean and standard appendicitis (630), Class 1: appendicitis; [10] auto-regressive model deviation of acute appendicitis (970), 10-fold Class 2: acute appendicitis; 81.3 80.3 80.8 and linear kernel prediction error duodenitis (1380) and Class 3: duodenitis; SVM pancreatitis (820) Class 4: pancreatitis peak systolic velocity; reverse velocity; peak Healthy (1800), Class 0: healthy; diastolic velocity; end appendicitis (630), Class 1: appendicitis; Binary Classiﬁcation [11] diastolic velocity; acute appendicitis (970), 10-fold Class 2: acute appendicitis; 81.7 82.9 82.3 using RBF SVM duration of systole; duodenitis (1380) and Class 3: duodenitis; and duration of pancreatitis (820) Class 4: pancreatitis diastole Healthy (1800), Class 0: healthy; spatial features A recursive cluster appendicitis (630), Class 1: appendicitis; obtained from a [32] elimination based acute appendicitis (970), 10-fold Class 2: acute appendicitis; 84.7 84.1 84.4 bi-modal Gaussian SVM duodenitis (1380) and Class 3: duodenitis; model pancreatitis (820) Class 4: pancreatitis Healthy (1800), Class 0: healthy; Periodic and appendicitis (630), Class 1: appendicitis; [14] RBF SVM non-periodic feature acute appendicitis (970), 10-fold Class 2: acute appendicitis; 85.3 86.1 85.7 extension duodenitis (1380) and Class 3: duodenitis; pancreatitis (820) Class 4: pancreatitis Healthy (1800), Class 0: healthy; Multinomial Cross-correlation and appendicitis (630), Class 1: appendicitis; Our work Classiﬁcation using convolution acute appendicitis (970), 10-fold Class 2: acute appendicitis; 92.0 91.2 91.6 customized kernel coecients duodenitis (1380) and Class 3: duodenitis; pancreatitis (820) Class 4: pancreatitis Appl. Sci. 2019, 9, 2284 13 of 14 5. Conclusions In this paper, a novel MOGA-SVM has been proposed for the multinomial classiﬁcation of four common organ inﬂammations, appendicitis, acute appendicitis, duodenitis, and pancreatitis. A customized similarity kernel K is optimally designed using MOGA. K captures the characteristics CS CS of the inﬂammations, which is an ideal approach in the kernel selection perspective. Typical kernel functions are generally built-in package as the analytic tool that does not aim at yielding best performance for all applications, and it is thus highly recommended that the customized kernel should be utilized for organ inﬂammations classiﬁcation. The results show that the proposed algorithm achieves sensitivity, speciﬁcity, and accuracy of 92%, 91.2%, and 91.6%, respectively. It achieves a signiﬁcant improvement using traditional kernels and related works by 60% and 10%, respectively. It is believed that WPS can be utilized as alternative, reliable and accurate method to determine whether a candidate is suering from organ inﬂammation. Besides accuracy, the proposed method is a timely and inexpensive approach. Bringing machine learning into real-world healthcare application is always a good solution to relieve the workload of medical personnel, as everybody needs regular body check and timely examination. Author Contributions: K.T.C. proposed and implemented the methodology. K.T.C. and M.D.L. have evaluated the performance of proposed work and drafted the paper. Funding: Authors would like to thank Eat University in Jeddah, Saudi Arabia, for funding the research reported in this paper through the Research and Consultancy Institute. Acknowledgments: Gratitude is expressed to Y.C., L.Z., D.Z. and D.Z., Department of Computing, Biometrics Research Center, The Hong Kong Polytechnic University, for sharing the datasets. Conﬂicts of Interest: The authors declare no conﬂict of interest. References 1. WHO. Health Workforce Requirements for Universal Health Coverage and the Sustainable Development Goals; World Health Organization: Geneva, Switzerland, 2016. 2. Chui, K.T.; Alhalabi, W.; Pang, S.S.H.; Pablos, P.O.D.; Liu, R.W.; Zhao, M. Disease Diagnosis in Smart Healthcare: Innovation, Technologies and Applications. Sustainability 2017, 9, 2309. [CrossRef] 3. Spruit, M.; Lytras, M. Applied Data Science in Patient-centric Healthcare. Telemat. Inform. 2018, 35, 643–653. [CrossRef] 4. Khaire, N.N.; Joshi, Y.V. Disgnosis of disease using wrist pulse signal for classiﬁcation of pre-meal and post-meal samples. In Proceedings of the 2015 International Conference on Industrial Instrumentation and Control, Maharashtra, India, 28–30 May 2015; IEEE: Piscataway, NJ, USA. 5. Reddy, R.K.; Pooni, R.; Zaharieva, D.P.; Senf, B.; El Youssef, J.; Dassau, E.; Castle, J.R. Accuracy of Wrist-Worn Activity Monitors during Common Daily Physical Activities and Types of Structured Exercise: Evaluation Study. JMIR mHealth uHealth 2018, 6, e10338. [CrossRef] [PubMed] 6. Li, J.; Zhang, B.; Lu, G.; You, J.; Zhang, D. Body surface feature-based multi-modal Learning for Diabetes Mellitus detection. Inf. Sci. 2019, 472, 1–14. [CrossRef] 7. He, D.; Wang, L.; Fan, X.; Yao, Y.; Geng, N.; Sun, Y.; Xu, L.; Qian, W. A new mathematical model of wrist pulse waveforms characterizes patients with cardiovascular disease—A pilot study. Med. Eng. Phys. 2017, 48, 142–149. [CrossRef] [PubMed] 8. Qiao, L.J.; Qi, Z.; Tu, L.P.; Zhang, Y.H.; Zhu, L.P.; Xu, J.T.; Zhang, Z.F. The Association of Radial Artery Pulse Wave Variables with the Pulse Wave Velocity and Echocardiographic Parameters in Hypertension. Evid. Based Complement. Altern. Med. 2018, 2018, 5291759. [CrossRef] 9. Zhang, Z.; Zhang, Y.; Yao, L.; Song, H.; Kos, A. A sensor-based wrist pulse signal processing and lung cancer recognition. J. Biomed. Inform. 2018, 79, 107–116. [CrossRef] [PubMed] 10. Chen, Y.; Zhang, L.; Zhang, D.; Zhang, D. Computerized wrist pulse signal diagnosis using modiﬁed auto-regressive models. J. Med. Syst. 2011, 35, 321–328. [CrossRef] Appl. Sci. 2019, 9, 2284 14 of 14 11. Chow, W.H.; Wu, C.K.; Tsang, K.F.; Li, B.Y.S.; Chui, K.T. Wrist pulse signal classiﬁcation for inﬂammation of appendix, pancreas, and duodenum. In Proceedings of the 40th Annual Conference of the IEEE Industrial Electronics Society, Dallas, TX, USA, 30 October–1 November 2014; IEEE: Piscataway, NJ, USA. 12. Garg, N.; Bisht, A.; Ryait, H.S.; Kumar, A. Identiﬁcation of motion outliers in wrist pulse signal. Comput. Electr. Eng. 2018, 67, 776–790. [CrossRef] 13. Liu, X.; Ji, Z.; Tang, Y. Recognition of pulse wave feature points and non-invasive blood pressure measurement. J. Signal Process. Syst. 2017, 87, 241–248. [CrossRef] 14. Wang, D.; Zhang, D.; Lu, G. Generalized Feature Extraction for Wrist Pulse Analysis: From 1-D Time Series to 2-D Matrix. IEEE J. Biomed. Health Inform. 2017, 21, 978–985. [CrossRef] [PubMed] 15. WHO. Global Health Estimates 2015: Deaths by Cause, Age, Sex, by Country and by Region, 2000–2015; World Health Organization: Geneva, Switzerland, 2016. 16. Chui, K.T.; Tsang, K.F.; Chi, H.R.; Ling, B.W.K.; Wu, C.K. An accurate ECG based transportation safety drowsiness detection scheme. IEEE Trans. Ind. Inform. 2016, 12, 1438–1452. [CrossRef] 17. Montazeri, A.; West, C.; Monk, S.D.; Taylor, C.J. Dynamic modelling and parameter estimation of a hydraulic robot manipulator using a multi-objective genetic algorithm. Int. J. Control 2017, 90, 661–683. [CrossRef] 18. Tseng, F.H.; Wang, X.; Chou, L.D.; Chao, H.C.; Leung, V.C. Dynamic Resource Prediction and Allocation for Cloud Data Center Using the Multiobjective Genetic Algorithm. IEEE Syst. J. 2018, 12, 1688–1699. [CrossRef] 19. Chui, K.T.; Tsang, K.F.; Wu, C.K.; Hung, F.H.; Chi, H.R.; Chung, H.S.H.; Man, K.F.; Ko, K.T. Cardiovascular diseases identiﬁcation using electrocardiogram health identiﬁer based on multiple criteria decision making. Expert Syst. Appl. 2015, 42, 5684–5695. [CrossRef] 20. Fuchida, M.; Pathmakumar, T.; Mohan, R.E.; Tan, N.; Nakamura, A. Vision-based perception and classiﬁcation of mosquitoes using support vector machine. Appl. Sci. 2017, 7, 51. [CrossRef] 21. Wu, J.L.; Chang, P.C.; Tsao, C.C.; Fan, C.Y. A patent quality analysis and classiﬁcation system using self-organizing maps with support vector machine. Appl. Soft Comput. 2016, 41, 305–316. [CrossRef] 22. Ryan, W.L. Digestive Diseases—Research and Clinical Developments: Appendicitis: Symptoms, Diagnosis, and Treatments; Nova Science: New York, NY, USA, 2011. 23. Keyzer, C.; Gevenois, P.A. Imaging of Acute Appendicitis in Adults and Children; Springer: Berlin, Germany, 24. Serra, S.; Jani, P.A. An approach to duodenal biopsies. J. Clin. Pathol. 2006, 59, 1133–1150. [CrossRef] 25. Adams, D.B.; Cotton, P.B.; Zyromski, N.J.; Windsor, J. Pancreatitis: Medical and Surgical Management; Wiley Blackwell: Chichester, UK, 2017. 26. Yadav, D.; Timmons, L.; Benson, J.T.; Dierkhising, R.A.; Chari, S.T. Incidence, prevalence, and survival of chronic pancreatitis: A population-based study. Am. J. Gastroenterol. 2011, 106, 2192. [CrossRef] 27. De Haan, R.R.; Visser, J.B.; Pons, E.; Feelders, R.A.; Kaymak, U.; Hunink, M.M.; Visser, J.J. Patient-speciﬁc workup of adrenal incidentalomas. Eur. J. Radiol. Open 2017, 4, 108–114. [CrossRef] 28. Roberts, D.R.; Bahn, V.; Ciuti, S.; Boyce, M.S.; Elith, J.; Guillera-Arroita, G.; Hauenstein, S.; Lahoz-Monfort, J.J.; Schroder, B.S.; Thuiller, W.; et al. Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure. Ecography 2017, 40, 913–929. [CrossRef] 29. Deb, K. Multi-Objective Optimization Using Evolutionary Algorithms; John Wiley & Sons, Inc.: New York, NY, USA, 2001. 30. Herbrich, R. Learning Kernel Classiﬁers Theory and Algorithms; The MIT Press: London, UK, 2002. 31. Smits, G.F.; Jordan, E.M. Improved SVM regression using mixtures of kernels. In Proceedings of the 2002 International Joint Conference on Neural Networks, Honolulu, HI, USA, 12–17 May 2002; IEEE: Piscataway, NJ, USA. 32. Rangaprakash, D.; Dutt, D.N. Study of wrist pulse signals using time domain spatial features. Comput. Electr. Eng. 2015, 45, 100–107. [CrossRef] © 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/). http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Applied Sciences Multidisciplinary Digital Publishing Institute http://www.deepdyve.com/lp/multidisciplinary-digital-publishing-institute/a-novel-moga-svm-multinomial-classification-for-organ-inflammation-xkPYKJP402

Loading next page...

References (33)

Kwok Chui, K. Tsang, C. Wu, Faan Hung, H. Chi, H. Chung, Kim-Fung Man, K. Ko (2015)
Cardiovascular diseases identification using electrocardiogram health identifier based on multiple criteria decision making
Expert Syst. Appl., 42
Narendra Khaire, Y. Joshi (2015)
Diagnosis of Disease Using Wrist Pulse Signal for classification of pre-meal and post-meal samples
2015 International Conference on Industrial Instrumentation and Control (ICIC)
K. Deb (2001)
Multi-objective optimization using evolutionary algorithms
G. Smits, E. Jordaan (2002)
Improved SVM regression using mixtures of kernels
Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290), 3
Kwok Chui, K. Tsang, H. Chi, B. Ling, C. Wu (2016)
An Accurate ECG-Based Transportation Safety Drowsiness Detection Scheme
IEEE Transactions on Industrial Informatics, 12
J. Shamblin (1966)
Pancreatitis: medical and surgical management.
Journal of the Medical Association of the State of Alabama, 35 12
X. Liu, Zhong Ji, Yuran Tang (2017)
Recognition of Pulse Wave Feature Points and Non-invasive Blood Pressure Measurement
Journal of Signal Processing Systems, 87
Masataka Fuchida, T. Pathmakumar, R. Mohan, Ning Tan, Akio Nakamura (2017)
Vision-Based Perception and Classification of Mosquitoes Using Support Vector Machine
Applied Sciences, 7
D. Rangaprakash, D. Dutt (2015)
Study of wrist pulse signals using time domain spatial features
Comput. Electr. Eng., 45
G. Beluffi (2013)
Imaging of acute appendicitis in adults and children
La radiologia medica, 118
(2011)
Digestive Diseases-Research and Clinical Developments: Appendicitis: Symptoms, Diagnosis, and Treatments
Romy Haan, J. Visser, E. Pons, R. Feelders, U. Kaymak, M. Hunink, J. Visser (2017)
Patient-specific workup of adrenal incidentalomas
European Journal of Radiology Open, 4
Wai Chow, C. Wu, K. Tsang, B. Li, Kwok Chui (2014)
Wrist pulse signal classification for inflammation of appendix, pancreas, and duodenum
IECON 2014 - 40th Annual Conference of the IEEE Industrial Electronics Society
Jheng-Long Wu, P. Chang, Cheng-Chin Tsao, C. Fan (2016)
A patent quality analysis and classification system using self-organizing maps with support vector machine
Appl. Soft Comput., 41
Nidhi Garg, Amandeep Bisht, H. Ryait, Amod Kumar (2018)
Identification of motion outliers in wrist pulse signal
Comput. Electr. Eng., 67
Dianning He, Lu Wang, Xiaobing Fan, Yang Yao, N. Geng, Yingxian Sun, Lisheng Xu, Wei Qian (2017)
A new mathematical model of wrist pulse waveforms characterizes patients with cardiovascular disease - A pilot study.
Medical engineering & physics, 48
Yinghui Chen, Lei Zhang, David Zhang, Dongyu Zhang (2011)
Computerized Wrist Pulse Signal Diagnosis Using Modified Auto-Regressive Models
Journal of Medical Systems, 35
Fan-Hsun Tseng, Xiaofei Wang, Li-Der Chou, H. Chao, Victor Leung (2018)
Dynamic Resource Prediction and Allocation for Cloud Data Center Using the Multiobjective Genetic Algorithm
IEEE Systems Journal, 12
D. Roberts, V. Bahn, S. Ciuti, M. Boyce, J. Elith, G. Guillera‐Arroita, Severin Hauenstein, J. Lahoz‐Monfort, B. Schröder, W. Thuiller, D. Warton, B. Wintle, F. Hartig, C. Dormann (2017)
Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure
Ecography, 40
Jinxing Li, Bob Zhang, Guangming Lu, J. You, David Zhang (2019)
Body surface feature-based multi-modal Learning for Diabetes Mellitus detection
Inf. Sci., 472
S. Serra, Prashant Jani (2006)
An approach to duodenal biopsies
Journal of Clinical Pathology, 59
D. Yadav, Lawrence Timmons, J. Benson, R. Dierkhising, S. Chari (2011)
Incidence, Prevalence, and Survival of Chronic Pancreatitis: A Population-Based Study
The American Journal of Gastroenterology, 106
M. Spruit, Miltiadis Lytras (2018)
Applied data science in patient-centric healthcare: Adaptive analytic systems for empowering physicians and patients
Telematics Informatics, 35
Zhichao Zhang, Yuan Zhang, Lina Yao, Houbing Song, A. Kos (2018)
A sensor-based wrist pulse signal processing and lung cancer recognition
Journal of biomedical informatics, 79
(2011)
Digestive Diseases—Research and Clinical Developments: Appendicitis: Symptoms, Diagnosis, and Treatments; Nova Science
Dimin Wang, David Zhang, Guangming Lu (2017)
Generalized Feature Extraction for Wrist Pulse Analysis: From 1-D Time Series to 2-D Matrix
IEEE Journal of Biomedical and Health Informatics, 21
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license
R. Herbrich (2001)
Learning Kernel Classifiers: Theory and Algorithms
(2016)
Health Workforce Requirements for Universal Health Coverage and the Sustainable Development Goals; World Health Organization
A. Montazeri, C. West, S. Monk, C. Taylor (2017)
Dynamic modelling and parameter estimation of a hydraulic robot manipulator using a multi-objective genetic algorithm
International Journal of Control, 90
Li-Jie Qiao, Zhen Qi, Liping Tu, Yu-Hang Zhang, Li-Ping Zhu, Jia-tuo Xu, Zhi-Feng Zhang (2018)
The Association of Radial Artery Pulse Wave Variables with the Pulse Wave Velocity and Echocardiographic Parameters in Hypertension
Evidence-based Complementary and Alternative Medicine : eCAM, 2018
Kwok Chui, W. Alhalabi, Sally Pang, P. Pablos, R. Liu, Mingbo Zhao (2017)
Disease Diagnosis in Smart Healthcare: Innovation, Technologies and Applications
Sustainability, 9
Ravi Reddy, Rubin Pooni, D. Zaharieva, Brian Senf, J. Youssef, E. Dassau, F. III, M. Clements, M. Rickels, S. Patton, J. Castle, M. Riddell, P. Jacobs (2018)
Accuracy of Wrist-Worn Activity Monitors During Common Daily Physical Activities and Types of Structured Exercise: Evaluation Study
JMIR mHealth and uHealth, 6

Publisher: Multidisciplinary Digital Publishing Institute
Copyright: © 1996-2023 MDPI (Basel, Switzerland) unless otherwise stated Disclaimer Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. Terms and Conditions Privacy Policy
ISSN: 2076-3417
DOI: 10.3390/app9112284
Publisher site: See Article on Publisher Site

Abstract

applied sciences Article A Novel MOGA-SVM Multinomial Classiﬁcation for Organ Inﬂammation Detection 1 , 2 , 3 Kwok Tai Chui * and Miltiadis D. Lytras Department of Electronic Engineering, City University of Hong Kong, Hong Kong SAR, China School of Business & Economics, Deree College—The American College of Greece, 6 Gravias Street GR-153 42, Aghia Paraskevi, 15342 Athens, Greece; [email protected] Eat College of Engineering, Eat University, Jeddah P.O. Box 34689, Saudi Arabia * Correspondence: [email protected] Received: 6 April 2019; Accepted: 29 May 2019; Published: 3 June 2019 Featured Application: In this paper, a novel multi-objective genetic algorithm based support vector machine (MOGA-SVM) has been proposed. A customized similarity kernel has been optimally designed for the multinomial classiﬁcation of the inﬂammations of appendix, pancreas, and duodenum. Practically, this methodology can be applied to other classiﬁcation problems as the concept of the methodology is to customize the kernel to speciﬁc application. In order to achieve a better performance using kernel based algorithm, it is highly recommended to use customize kernel instead of traditional kernels. Abstract: Wrist pulse signal (WPS) contains crucial information of humans’ health condition. It can serve as an alternative method for diagnosing of organ inﬂammation instead of traditional clinical measurement. In this paper, a novel multi-objective genetic algorithm based support vector machine (MOGA-SVM) has been proposed for the multinomial classiﬁcation of the inﬂammations of appendix, pancreas, and duodenum. A customized similarity kernel (K ) has been optimally designed. The CS performance of multinomial classiﬁcation using K is compared with ﬁve types of kernels, linear, CS radial basis function (RBF), polynomial and sigmoid kernel, as well as mixtures of polynomial and RBF, to verify the eectiveness of K . The sensitivity, speciﬁcity and accuracy (Acc) of the proposed CS method are 92%, 91.2%, and 91.6% respectively. The results have demonstrated that K improves CS the accuracy of classiﬁcation from 8.9% to 59.6%. When compared to related work, the proposed method increases the performance by more than 10%. It is believed that WPS can serve as alternative measures to diagnose organ inﬂammations. Keywords: bioinformations; genetic algorithm; multiobjective optimization; organ inﬂammation; support vector machine; wrist pulse signal 1. Introduction Health is crucial element in today’s life. Researchers have devoted vast eorts in proposing new policies, algorithms, systems, and architectures for healthcare. According to the World Health Organization (WHO), in 2013, the global requirement and the actual number of health workforce were 60.4 million and 43 million, respectively [1]. These ﬁgures will be increased to 81.8 million and 67.3 million, respectively, by 2030. Hence, it is believed that the shortage of medical personnel is unsolved and remained serious in the coming decade. Automatic decision making via machine learning is believed to be the only way out to solve the shortage of medical personnel [2,3]. Medical workers may argue that the automatic system has a conﬂict of interest with them; nevertheless, it is not the truth. First, the current workload of medical workers (ratio of workers to patients) is heavy Appl. Sci. 2019, 9, 2284; doi:10.3390/app9112284 www.mdpi.com/journal/applsci Appl. Sci. 2019, 9, 2284 2 of 14 and will become normal. Second, an automatic system focuses on routine works, so that medical workers can devote more time to professional consultation and surgery activities. Third, the increase in quality of medical services will lead to higher acceptance and satisfaction by the public. Thus, medical workers will earn a higher social status and better job satisfaction. Many diseases and abnormal human conditions can be examined by digital imaging diagnostic, like X-ray, Computed Tomography (CT) scan, Magnetic Resonance Imaging (MRI), Ultrasonography, Electrocardiogram, and Biopsy. In this paper, Wrist pulse signal (WPS) of human is considered which provides key information regarding health conditions. In the literature, WPS can be utilized for various applications, for instance, pre-meal and post-meal classiﬁcation [4], physical exercise [5], diabetes classiﬁcation [6], hypertension association [7,8], lung cancer recognition [9], and inﬂammation classiﬁcation [10,11]. Various signal processing techniques on WPS can be found in [12–14], for instance, dynamic time warping, wavelet analysis, periodic decomposition, principal component analysis, and linear discriminant analysis. In this paper, four common types of organ inﬂammation are considered, namely, appendicitis, acute appendicitis, duodenitis, and pancreatitis. According to the WHO, the annual deaths hat are attributable to appendicitis, duodentitis and pancreatitis in 2000, 2005, 2010, and 2015 are shown in Table 1 [15]. It is noted that acute appendicitis is embedded into Global Health Expand (GHE) code 1240. From Table 1, the number of deaths in each category is increasing by an increment of29%, 24%, and 60% for appendicitis, duodenitis, and pancreatitis, respectively, from 2000 to 2015. Among three types of organ inﬂammations, pancreatitis is the leading cause, which is followed by duodenitis and appendicitis. To conclude, the issues of deaths in these organ inﬂammations remain unsolved. Table 1. Number of deaths due to Appendicitis, Pancreatitis, and Duodenitis in 2000, 2005, 2010, and 2015. Number of Deaths (Annual) GHE Cause GHE Code 2000 2005 2010 2015 1240 Appendicitis 34,800 39,400 43,300 45,000 1241 Duodenitis 37,900 40,400 43,800 47,000 1248 Pancreatitis 64,400 77,800 93,900 103,500 There have been more than million of suerers and thus it is necessary to have a reliable and accurate method for the diagnosis of organ inﬂammations. Based on literature ﬁnding, there are a few publications working on binary classiﬁcation of healthy, appendicitis, acute appendicitis, duodenitis, and pancreatitis suerers [10,11]. In [10], the features extraction process, an auto-regression (AR) based model was proposed. Two features, the standard deviation and mean of the prediction error from AR model, were chosen to represent the information of the WPS, and for further analysis. With regard to the classiﬁcation, the support vector machine (SVM) with linear kernel was adopted for the binary classiﬁcation, which yields an accuracy of 77.8–91.2%. For further improvement, a radial basis kernel (RBF) has been utilized to replace the linear kernel [11]. The idea is that most of the classiﬁcation problems are not linearly separable. The enhanced method achieved an accuracy of 88.6–98.4%. Nevertheless, as a pragmatic application, it is deemed to be formulated as classifying instances into one of the more than two classes, and multinomial classiﬁcation is desired. A novel multi-objective genetic algorithm (MOGA) based SVM, abbreviated as MOGA-SVM, has been proposed for the multinomial classiﬁcation of the organ inﬂammations of appendicitis, acute appendicitis, duodenitis, and pancreatitis. MOGA is a heuristic approach that has been widely adopted to obtain tradeo solutions between two or more conﬂicting objectives [16–18]. SVM receives a lot of attention as a supervised learning algorithm for classiﬁcation problems [19–21]. In this paper, a customized similarity kernel (K ) has been optimally designed for speciﬁc application, the classiﬁcation of appendicitis, CS acute appendicitis, duodenitis, and pancreatitis. It is worth mentioning that traditional kernels, like linear, RBF, quadratic, and polynomial kernels are not designed for any particular application. It is Appl. Sci. 2019, 9, 2284 3 of 14 recommended that the customized kernel should be utilized for organ inﬂammations classiﬁcation instead of traditional kernels in order to improve the classiﬁcation accuracy. This paper is organized, as follows. Section 2 provides the background of organ inﬂammations and an overview of MOGA-SVM. The methodology of the proposed algorithm is explained in Section 3. Performance evaluation and comparison are given in Section 4. Finally, a conclusion is made in Section 5. 2. Dataset and Overview of MOGA-SVM The background symptoms of each organ inﬂammation, appendicitis, acute appendicitis, duodenitis, and pancreatitis will ﬁrstly describe. Only a summary is provided in each topic. Readers who are interested in the details of the inﬂammations are suggested to refer to appendicitis [22], acute appendicitis [23], duodenitis [24], and pancreatitis [25]. Subsequently, the overview of the MOGA-SVM is brieﬂy discussed, in which the details will be explained in the next section. 2.1. Background of Organ Inﬂammation 2.1.1. Appendicitis Appendicitis is an inﬂammation of the appendix. It is not uncommon abdominal emergency at any age. The causes are due to the blocking of appendix by stool, cancer, or foreign body, or from infection. Typical symptoms are abdominal pain, loss of appetite, diarrhea, and inability to pass gas. The clinical approaches for diagnosis include abdominal exam, urine test, rectal exam, blood test, CT scans, and ultrasound. The lifetime risk of suering from appendicitis is about 7%, with dierent severity levels [22]. The occurrence of this inﬂammation is approximately 11 persons per 10,000 populations in each year. 2.1.2. Acute Appendicitis Acute appendicitis is more severe than appendicitis, which has annual incidence of 90 to 140 per 10,000 populations [23]. Although this inﬂammation has been documented for more than 500 years, its etiology is not well known. It is usually results from injury of its mucosa and spread from that injury via its wall. The symptoms and examinations of acute appendicitis are similar to that in appendicitis. 2.1.3. Duodenitis Duodenitis is inﬂammation of the duodenum. The known causes include helicobacter pylori infection, bacterial infection, Nonsteroidal anti-inﬂammatory drug, viral infection, coeliac disease, and idiopathic [24]. Abdominal pain, nausea, vomiting, and discomfort in stomach are the four known symptoms. The most common examination is an Oesophago-Gastro-Duodenoscopy. For the global annual years of healthy life lost, the estimation is about 58 persons per 100,000. 2.1.4. Pancreatitis Pancreatitis is inﬂammation of the pancreas. It is more important than the aforementioned three organ inﬂammations, because it often characterized by irreversible change, permanent loss of function [25]. The clinical features of pancreatitis include ﬁbrosis, chronic and recurrent inﬂammation, duct distortion, atrophy, and the risk of pancreatic cancer. The estimated incidence of pancreatitis is 42 persons per 100,000 population [26]. 2.2. Overview of MOGA-SVM Figure 1 shows the ﬂow chart of MOGA-SVM for organ inﬂammations classiﬁcation. The typical waveform of the wrist pulse signal is shown in Figure 2, which is characterized by a percussion wave, tidal wave, dicrotic wave, peak systolic velocity, reverse velocity, peak diastolic velocity, and end diastolic velocity. Appl. Sci. 2019, 9, x 5 of 15 Appl. Sci. 2019, 9, 2284 4 of 14 Appl. Sci. 2019, 9, x 5 of 15 Figure 1. Overview of multi-objective genetic algorithm based support vector machine (MOGA-SVM) Figure 1. Overview of multi-objective genetic algorithm based support vector machine (MOGA-SVM) Figure 1. Overview of multi-objective genetic algorithm based support vector machine (MOGA-SVM) for organ inflammations classification. for organ inflammations classification. for organ inﬂammations classiﬁcation. Figure 2. Typical waveform of wrist pulse signal. The similarity coefficients of every pair of WPS are computed, which form the customized Figure 2. Typical waveform of wrist pulse signal. Figure 2. Typical waveform of wrist pulse signal. similarity kernel. After MOGA, the optimal kernel KCS is designed. A classifier for organ inflammations classification is constructed. Section 3discusses the details (Figure 3 is drawn to The training of organ inﬂammations classiﬁer, the datasets contain WPS of healthy, appendicitis, The similarity coefficients of every pair of WPS are computed, which form the customized summarize the key steps of the MOGA-SVM). In this paper, the 10-fold cross-validation is adopted acute appendicitis, duodenitis, and pancreatitis candidates were considered [10]. Each record of WPS similarity kernel. After MOGA, the optimal kernel KCS is designed. A classifier for organ to evaluate the classifier, as it is a practical order in literature [27,28]. is carried out DC drift elimination and low-pass ﬁlter following the approach, as in [11]. Afterwards, inflammations classification is constructed. Section 3discusses the details (Figure 3 is drawn to the local maxima and minima points of the WPS are located. The detail is not being discussed in this summarize the key steps of the MOGA-SVM). In this paper, the 10-fold cross-validation is adopted paper, as the authors would like to mainly focus on the proposed MOGA-SVM. to evaluate the classifier, as it is a practical order in literature [27,28]. Appl. Sci. 2019, 9, 2284 5 of 14 The similarity coecients of every pair of WPS are computed, which form the customized similarity kernel. After MOGA, the optimal kernel K is designed. A classiﬁer for organ inﬂammations CS Appl. Sci. 2019, 9, x 9 of 15 classiﬁcation is constructed. Section 3 discusses the details (Figure 3 is drawn to summarize the key steps of the MOGA-SVM). In this paper, the 10-fold cross-validation is adopted to evaluate the classiﬁer, generations or if the output reaches the pareto front. To facilitate readers’ understanding, Algorithm as it is a practical order in literature [27,28]. 1 and Algorithm 2 are presented as the pseudo code of MOGA-SVM. Figure 3. Optimal design of KCS and classifiers using MOGA-SVM. Figure 3. Optimal design of K and classiﬁers using MOGA-SVM. CS A pseudo code for the MOGA-SVM is given for better illustration and understanding. When it comes to practical application, the WPS of the candidate is measured and it served as the Algorithm 1 Segmentati on ( X ) input of the trained organ inﬂammations classiﬁer. The outputs maintain ﬁve possibilities, healthy, appendicitis, acute appendicitis, duodenitis, and pancreatitis. If the status is one of the four organ Data: Organ inflammations of appendicitis, acute appendicitis, duodenitis and pancreatitis retrieved inﬂammations, from 248 cand aid rat eport es [1will 0], Xbe m sent to a nurse and doctor for further examination and treatment. Output: WPS samples Xi,j 3. Methodology Step 1: dc drift elimination Step 2: Filter Xm using low pass filter Hlow This section is composed of three parts. First, the datasets of healthy, appendicitis, acute Step 3: Locate local maxima and minima points of the Xm; appendicitis, duodenitis, and pancreatitis candidates are illustrated in Section 3.1. Next, the data Step 4: Locate two maxima points with interval of 120 sampling points; preprocessing of the datasets is explained in Section 3.2. At last, Section 3.3 formulates the optimal Xi,j (i=1:4=class label, j=length(Class))←Portion of signal between two maxima points with interval of design of the K . CS 120 sampling points 3.1. Datasets of Organ Inﬂammations Classiﬁer Algorithm 2TrainClass ifier (Classlabel , K , K ) c cc Gratitude is expressed to researchers in [10] for sharing the datasets. The WPSs were measured Data: Classlabel, Kc, Kcc while using Doppler ultrasonic blood analyzer module. In each measurement, three steps were Output: Model followed. Firstly, an approximated position was located where the ﬂuctuation of signal was larger Step 1: generations = 1 than the other positions. Subsequently, a ﬁne tuning of position with slight variation of angle and Step 2: initialization (population) position until the largest signal was observed. Finally, WPS was recorded under the setting of largest Step 3: Evaluate the individuals with the fitness function (F1 and F2) signal amplitude. Step 4: rank the individuals by their fitness values by step 3 Table 2 summarizes the details of the datasets. Assign the class label to each of the category, Step 5: do the Niche count calculation Class 0: healthy, Class 1: appendicitis, Class 2: acute appendicitis, Class 3: duodenitis, and Class 4: while generations <= max_generationdo pancreatitis. The datasets are formed by four age groups, [0, 20), [20, 40), [40, 60) and [60, 100). The Step 6: Select two parents from the population total number of samples is 248 and the corresponding samples in Class 0–4 are 100, 22, 38, 42, and Step 7: Create the offspring using Roulette wheel selection,crossover and mutation 46, respectively. Step 8: Train SVM model for each individual Step 9: Evaluate the offspring with the fitness function (F1and F2) Step 10: rank the individuals by their fitness values by step 3 Step 11: do the Niche count calculation Appl. Sci. 2019, 9, 2284 6 of 14 Table 2. Sample distribution of the datasets. Age Class Name Total [0,20) [20,40) [40,60) [60,100) 0 Healthy 8 26 30 16 100 1 Appendicitis 0 22 0 0 22 2 Acute Appendicitis 20 8 10 0 38 3 Duodenitis 4 26 6 6 42 4 Pancreatitis 16 26 4 0 46 3.2. Data Preprocessing The data preprocessing of the aforementioned samples is following the related work [11]. It includes DC drift elimination, six-order Butterworth low-pass ﬁlter, and the detection of local maxima and minima points. In this analysis, the WPS has a cycle less than 120 samples. The individual sample is formed by the portion between the two largest maxima points. There are 1800, 630, 972, 1386, and 828 samples for healthy, appendicitis, acute appendicitis, duodenitis, and pancreatitis candidates, respectively. For equal division using 10-fold cross validation, two, six, and eight samples have been removed for acute appendicitis, duodenitis, and pancreatitis candidates. Overall, there are 5600 samples. 3.3. Formulation of Optimal K and MOGA-SVM Classiﬁer CS Kernel is essential in SVM classiﬁcation and it has to obey Mercer ’s theorem. That is, the kernel is positive semi-deﬁnite. A common interpretation of kernel is that it captures the correlation between pairs of data. Thus, the proposed K is optimally designed using convolution and cross-correlation. CS The K is formulated as multi-objective optimization problem and is solved by MOGA [29]. CS Let X (n) of length 120 (zero padding for length<120) be the WPS sample. The subscript i refers to i,j the class label from 0 to 4 and that of j refers to the sample number. Therefore, the sets in Class 0 to Class 4 are {X (n), ::: ,X (n)}, {X (n), ::: ,X (n)},{X (n), ::: ,X (n)},{X (n), ::: ,X (n)},and 0,1 0,1800 1,1 1,630 2,1 2,970 3,1 3,1380 {X (n), ::: ,X (n)}, respectively. 4,1 4,820 The convolution between two WPSs X (n) and X (n) is given by a,b c,d N1 a,b C (n) = X (n) X (n) = X (k)X (n k) (1) a,b c,d a,b c,d c,d k=0 where N = 120 is the length of the WPS sample. The cross-correlation between two WPSs X (n) and X (n) can be expressed as a,b c,d N1 X (n)X (n k), k 0 > a,b c,d a,b n=k R (k) = (2) > Njkj1 c,d X (n)X (n k), k < 0 a,b c,d n=0 The customized similarity kernel K is formulated by customized convolution kernel K and CS c customized cross-correlation kernel K . K , and K are deﬁned as cc c cc 2 3 6 X X 7 c,1,1 c,1,N 6 t 7 6 7 6 7 6 . . 7 6 7 . . . K = (3) 6 7 c . 6 . . 7 6 7 6 7 4 5 X X c,N ,N c,N ,1 t t t Appl. Sci. 2019, 9, 2284 7 of 14 2 3 6 X X 7 cc,1,1 cc,1,N 6 t 7 6 7 6 7 6 . . 7 6 7 . . . K = (4) 6 7 cc . 6 . . 7 6 7 6 7 4 5 X X cc,N ,1 cc,N ,N t t where N = 5040 is the 90% of the training samples in Class 0 to Class 4. Here, X refers to the t c,i,j weighting sum of convolution coecients between ith and jth WPS sample. The 1st to 1620th samples come from Class 0. The 1621th to 2187th samples come from Class 1. The 2188th to 3060th samples come from Class 2. The 3061th to 4302th samples come from Class 3. The 4303th to 5040th samples come from Class 4. X and X are computed by c,i,j cc,i,j 2N1 a,b X = w C (m) (5) c,i, j c,m c,d m=1 2N1 a,b X = w R (m) (6) cc,i, j cc,m c,d m=1 where w and w are the weightings of convolution coecients and cross-correlation coecients, c,m cc,m respectively. These weightings are optimally designed while using MOGA. The kernels K and K predominantly aect the maximum margin and the accuracy (Acc) of the c cc organ inﬂammations classiﬁer. From (3) and (4), the maximum margin is directly related to both X ,i,j and X . Based on (5) and (6), an optimal design of both X and X can be obtained by an optimal cc,i,j x,i,j cc,i,j a,b design of the weighting sw and w form = 1, ::: ,2N1, for the given sequences of C (m) and c,m cc,m c,d a,b R (m), respectively. In general, an optimally designed kernel will speed up the convergence of the c,d training algorithm MOGA-SVM. However, varying the combinations of w and w will deduce the c,m cc,m dierent kernels. Searching the optimal weightings require a large computational power. As a result, there is a tradeo between the accuracy and computational power. As it is dicult to ﬁnd the optimal values of w and w that are attributable to the complexity of the objective function, a good trial of c,m cc,m w and w are primarily important, which determines the accuracy. c,m cc,m In this paper, an multi-objective optimization approach, MOGA, is employed to determine the weighting sw and w . However, exhaustive search algorithms may not be the appropriate choices c,m cc,m for searching the solution of the optimization problem. The reason is that it deals with a huge range of combinations. Indeed, heuristic search algorithms eciently and eectively perform the searching of the optimal solutions. In particular, the GA is a robust searching heuristic algorithm that imitates the process of the natural evolution for searching the solution of the optimization problem by the operations in selection, inheritance, crossover and mutation. Aforementioned, to be a proper kernel for SVM classification, the fulfillment of Mercer ’s theorem is essential [30]. The evaluation of eigenvalues helps to determine the positive semi-definite of K and K . c cc K V = D V c c c c (7) K V = D V cc cc cc cc where V and V are non-zero eigenvectors for K and K , respectively. D and D are the c cc c cc c cc corresponding eigenvalues. All of the eigenvalues must be positive in order to ensure K and K are positive semi-deﬁnite. cc Deﬁne K as the sum of K and K . It takes the advantageous from K and K . It is proved below cs c cc c cc that the sum of Mercer ’s kernels is also a Mercer ’s kernel. If K and K are positive semi-deﬁnite, then c cc n T T for any c 2 < , c K c 0 and c K c 0. Hence, c cc T T c K c = c (K + K )c (8) CS c cc Appl. Sci. 2019, 9, 2284 8 of 14 T T T c K c = c K c + c K c 0 (9) CS c cc Therefore, the K is positive semi-deﬁnite. Thus, it is a Mercer ’s kernel. The multi objective CS optimization problem is formulated with two objective functions: Max F = M(, w) (10) Max F = OA = 0.5(S + S ) 2 e p > 0, y = 0, i = 1,::: , N i i i > i=1 2N1 2N1 P P s.t.> (11) > w = 1, w = 1 c,n cc,n n=1 n=1 D 0, D 0,8i c,i cc,i where M(, w) is the margin function of the classiﬁer, is the Lagrange multiplier, S is the sensitivity, i e S is the speciﬁcity, y 2 f1, +1g is the output of the classiﬁer, and D and D are the entries of D p i c,i cc,i c and D , respectively. The margin function is deﬁned as cc N N N X X X M(, w) = y y K (12) i i j i j CS i=1 i=1 j=1 Figure 3 shows the MOGA-SVM for the optimal design of the K . The procedures are as follows: CS (i) The population size and values of objective function are initialized; (ii) The values of objective function of individuals in the population are computed while using the values of objective function deﬁned in (i); (iii) Ranking the individuals according to the values of objective function; (iv) The population convergence is dependent on small group of pareto optimal solutions, but not all optimal solutions are attributable to the nature of the stochastic selection errors, given a limited population size; (v) Niche count is introduced to enhance the population diversity by lengthening the distance between two optimal solutions along the axis of objective functions. The convergence to small group solutions will be avoided; (vi) A new ospring is generated and the values of the objective functions are evaluated; (vii) Ranks assignment and niche count calculation are carried out repeatedly in the new ospring; and, (viii) The algorithm is terminated if it attains the maximum number of generations or if the output reaches the pareto front. To facilitate readers’ understanding, Algorithm 1 and Algorithm 2 are presented as the pseudo code of MOGA-SVM. A pseudo code for the MOGA-SVM is given for better illustration and understanding. Algorithm 1 Segmentation(X Data: Organ inﬂammations of appendicitis, acute appendicitis, duodenitis and pancreatitis retrieved from 248 candidates [10], X Output: WPS samples X i,j Step 1: dc drift elimination Step 2: Filter X using low pass ﬁlter H m low Step 3: Locate local maxima and minima points of the X ; Step 4: Locate two maxima points with interval of 120 sampling points; X (i = 1:4 = class label, j = length(Class)) Portion of signal between two maxima points with interval of 120 i,j sampling points Appl. Sci. 2019, 9, 2284 9 of 14 Algorithm 2 TrainClass i f ier(Classlabel, K , K ) c cc Data: Classlabel, K , K c cc Output: Model Step 1: generations = 1 Step 2: initialization (population) Step 3: Evaluate the individuals with the ﬁtness function (F1 and F2) Step 4: rank the individuals by their ﬁtness values by step 3 Step 5: do the Niche count calculation while generations <= max_generation do Step 6: Select two parents from the population Step 7: Create the ospring using Roulette wheel selection,crossover and mutation Step 8: Train SVM model for each individual Step 9: Evaluate the ospring with the ﬁtness function (F1and F2) Step 10: rank the individuals by their ﬁtness values by step 3 Step 11: do the Niche count calculation Step 12: Decide the new population based on the ospring Step 13: generations = generations + 1 End while Model Pareto solutions 4. Performance Evaluation and Comparison Section 4 is divided into three sub-sections. Firstly, the performance of the proposed K is CS analyzed. Afterwards, it is compared with ﬁve other kernels using the feature extraction approach. Finally, performance comparison between proposed and related work is discussed. 4.1. Performance of Proposed MOGA-SVM Using K CS The performance evaluation of the proposed MOGA-SVM using K adopts 10-fold cross CS validation. Randomly divide 5600WPS samples into 10 equal-sized subsets; with each set containing 560 (10%) samples with Class 0: 180 samples, Class 1: 63 samples, Class 2: 97 samples, Class 3: 138 samples, and Class 4: 82 samples. In each fold of validation, 90% of datasets (nine subsets) from each class serves as training dataset and 10% of the remaining subset serves as the testing datasets. This process completes one-fold of operations. Subsequently, another set is chosen for validation and the remaining nine subsets are used for training. It is noted that this chosen validation set must be dierent from the validation sets that were selected in the previous folds of operations. The process is repeated until all of the 10 subsets have been validation. Applying 10-fold cross validation, the proposed MOGA-SVM using KCS achieves average S , S , e p and Acc of 92%, 91.2%, and 91.6%, respectively. 4.2. Evaluation of Other Kernels Using Feature Extraction Approach In this subsection, feature extraction using convolution coecients and cross-correlation coecients as features is adopted. The following ﬁve kernels, linear, RBF polynomial and sigmoid kernel, and mixtures of polynomial and RBF kernels [31] are applied. They can be expressed by: D E Linear kernel : k (x , x ) = x , x (13) 1 i j i j RBF kernel : k (x , x ) = exp( x x /2) (14) 2 i j i j 2 D E Polynomial kernel : k (x , x ) = x , x + c (15) i j i j D E Sigmoid kernel : k (x , x ) = tanh x , x + c (16) 4 i j i j Mixtures of polynomial and RBF kernels k (x , x ) = k (x , x ) + (1 )k (x , x ) (17) 5 i j 3 i j 4 i j Appl. Sci. 2019, 9, 2284 10 of 14 Three scenarios are considered: (i) Only convolution coecients serve as features (1–199 coecients); (ii) Only cross-correlation coecients serve as features (1–199 coecients); and, (iii) Both convolution and cross-correlation coecients serve as features (1–398 coecients). Table 3 summarizes the performance of kernels K –K in three scenarios. Only the best scenario 1 5 is given. The results reveal that scenario (iii) achieves highest performance, because it takes the advantages from both the convolution and cross-correlation coecients. Compared K with K –K , CS 1 5 the ranking (from highest to lowest) is K > K > K > K > K > K . When compared to scenarios (i), CS 5 3 2 4 1 (ii), and (iii), K improves the Acc by 14.4–58.2%, 12.4–59.6%, and 8.9–53.7%, respectively. CS Table 3. Analysis of traditional kernels in organ inﬂammation classiﬁcations. Performance Kernel Scenario (i) Scenario (ii) Scenario (iii) (S ,S ,Acc)% (S ,S ,Acc)% (S ,S ,Acc)% e p e p e p k (x ,x ) (57.6, 58.2, 57.9) (57.7, 57.1, 57.4) (58.8, 60.4, 59.6) 1 i j k (x ,x ) (76.7, 77.5, 77.1) (76.8, 76.6, 76.7) (77.3, 78.3, 77.8) 2 i j k (x ,x ) (77.6, 78.2, 77.9) (78.3,78.9, 78.6) (78.7, 80.1, 79.4) 3 i j k (x ,x ) (73.8, 74.6, 74.2) (73.2, 73.0, 73.1) (74.8, 75.8, 75.3) 4 i j k (x ,x ) (79.9, 80.3, 80.1) (82.0, 81.0, 81.5) (83.8, 84.4, 84.1) 5 i j 4.3. Comparison between Proposed and Related Work Based on our ﬁnding, the multinomial classiﬁcation of appendicitis, acute appendicitis, duodenitis, and pancreatitis is the ﬁrst of its kind. Previous works [10,11] have considered the problem as binary classiﬁcation. To compare the performance between the proposed and related work [10,11], it is analyzed in two directions. (i) Table 4 gives the raw comparison between the works. (ii) Table 5 gives the matched comparison between the works. The forms of the datasets, application, and cross-validation in [10,11] will be changed into those in this paper. Thus, every work considers 5600 samples for multinomial classiﬁcation and evaluates using 10-fold cross validation. From the raw comparison, it can be seen that the performance, S , S , and Acc of classiﬁcation e p between [10] and [11] are similar for binary classiﬁcation between healthy and appendicitis candidates, and between healthy and duodenitis candidates. For that between healthy and acute appendicitis, and between healthy and pancreatitis, the improvements are 8% and 9%, respectively. If the proposed work is taken into account, it outperforms [10] in the classiﬁcation of all inﬂammations. By averaging the Acc in [11], it is approximately equal to the proposed work. Therefore, it can be interpreted that multinomial classiﬁcation can be achieved without deteriorating the performance in inﬂammations classiﬁcation. A matched comparison environment is setup to compare the performance between algorithms in organ inﬂammations classiﬁcation. Repeated simulation is carried out for [10,11,14,32] while using the identical datasets and 10-fold cross validation. It is concluded that the proposed MOGA-SVM improves the Acc from 6.9% to 13.4%. Appl. Sci. 2019, 9, 2284 11 of 14 Table 4. Raw comparison between proposed and related work [10,11]. Cross Work Method Feature Extraction Dataset (Samples) Class Labels S (%) S (%) Acc (%) e p Validation Class 0: healthy; 81.8 93.3 91.2 Class 1: appendicitis Class 0: healthy; Healthy (100), 76.5 82.4 80.8 Binary Classiﬁcation Class 1: acute appendicitis Mean and standard appendicitis (22), acute using modiﬁed Class 0: healthy; deviation of appendicitis (38), [10] No auto-regressive model 80.0 91.4 88.0 Class 1: duodenitis prediction error duodenitis (42) and and linear kernel pancreatitis (46) SVM Class 0: healthy; 83.3 94.4 90.9 Class 1: pancreatitics Class 0: healthy; 80.4 89.7 87.3 Class 1: All inﬂammations Class 0: healthy; N/A N/A 92.8 Class 1: appendicitis peak systolic velocity; Healthy (100), reverse velocity; peak Class 0: healthy; appendicitis (100), N/A N/A 88.1 diastolic velocity; end Class 1: acute appendicitis Binary Classiﬁcation [11] acute appendicitis (100), diastolic velocity; 10-fold using RBF SVM Class 0: healthy; duodenitis (100) and duration of systole; N/A N/A 88.6 Class 1: duodenitis pancreatitis (100) and duration of diastole Class 0: healthy; N/A N/A 98.4 Class 1: pancreatitics Healthy (1800), Class 0: health; Multinomial Cross-correlation and Appendicitis (630), Class 1: appendicitis; Our work Classiﬁcation using convolution Acute Appendicitis 10-fold Class 2: acute appendicitis; 92.0 91.2 91.6 customized kernel coecients (970), Duodenitis (1380) Class 3: duodenitis; and Pancreatitis (820) Class 4: pancreatitis Appl. Sci. 2019, 9, 2284 12 of 14 Table 5. Matched comparison between proposed and related work [10,11,32]. Cross Work Method Feature Extraction Dataset (Samples) Class Labels S (%) S (%) Acc (%) e p Validation Binary Classiﬁcation Healthy (1800), Class 0: healthy; using modiﬁed Mean and standard appendicitis (630), Class 1: appendicitis; [10] auto-regressive model deviation of acute appendicitis (970), 10-fold Class 2: acute appendicitis; 81.3 80.3 80.8 and linear kernel prediction error duodenitis (1380) and Class 3: duodenitis; SVM pancreatitis (820) Class 4: pancreatitis peak systolic velocity; reverse velocity; peak Healthy (1800), Class 0: healthy; diastolic velocity; end appendicitis (630), Class 1: appendicitis; Binary Classiﬁcation [11] diastolic velocity; acute appendicitis (970), 10-fold Class 2: acute appendicitis; 81.7 82.9 82.3 using RBF SVM duration of systole; duodenitis (1380) and Class 3: duodenitis; and duration of pancreatitis (820) Class 4: pancreatitis diastole Healthy (1800), Class 0: healthy; spatial features A recursive cluster appendicitis (630), Class 1: appendicitis; obtained from a [32] elimination based acute appendicitis (970), 10-fold Class 2: acute appendicitis; 84.7 84.1 84.4 bi-modal Gaussian SVM duodenitis (1380) and Class 3: duodenitis; model pancreatitis (820) Class 4: pancreatitis Healthy (1800), Class 0: healthy; Periodic and appendicitis (630), Class 1: appendicitis; [14] RBF SVM non-periodic feature acute appendicitis (970), 10-fold Class 2: acute appendicitis; 85.3 86.1 85.7 extension duodenitis (1380) and Class 3: duodenitis; pancreatitis (820) Class 4: pancreatitis Healthy (1800), Class 0: healthy; Multinomial Cross-correlation and appendicitis (630), Class 1: appendicitis; Our work Classiﬁcation using convolution acute appendicitis (970), 10-fold Class 2: acute appendicitis; 92.0 91.2 91.6 customized kernel coecients duodenitis (1380) and Class 3: duodenitis; pancreatitis (820) Class 4: pancreatitis Appl. Sci. 2019, 9, 2284 13 of 14 5. Conclusions In this paper, a novel MOGA-SVM has been proposed for the multinomial classiﬁcation of four common organ inﬂammations, appendicitis, acute appendicitis, duodenitis, and pancreatitis. A customized similarity kernel K is optimally designed using MOGA. K captures the characteristics CS CS of the inﬂammations, which is an ideal approach in the kernel selection perspective. Typical kernel functions are generally built-in package as the analytic tool that does not aim at yielding best performance for all applications, and it is thus highly recommended that the customized kernel should be utilized for organ inﬂammations classiﬁcation. The results show that the proposed algorithm achieves sensitivity, speciﬁcity, and accuracy of 92%, 91.2%, and 91.6%, respectively. It achieves a signiﬁcant improvement using traditional kernels and related works by 60% and 10%, respectively. It is believed that WPS can be utilized as alternative, reliable and accurate method to determine whether a candidate is suering from organ inﬂammation. Besides accuracy, the proposed method is a timely and inexpensive approach. Bringing machine learning into real-world healthcare application is always a good solution to relieve the workload of medical personnel, as everybody needs regular body check and timely examination. Author Contributions: K.T.C. proposed and implemented the methodology. K.T.C. and M.D.L. have evaluated the performance of proposed work and drafted the paper. Funding: Authors would like to thank Eat University in Jeddah, Saudi Arabia, for funding the research reported in this paper through the Research and Consultancy Institute. Acknowledgments: Gratitude is expressed to Y.C., L.Z., D.Z. and D.Z., Department of Computing, Biometrics Research Center, The Hong Kong Polytechnic University, for sharing the datasets. Conﬂicts of Interest: The authors declare no conﬂict of interest. References 1. WHO. Health Workforce Requirements for Universal Health Coverage and the Sustainable Development Goals; World Health Organization: Geneva, Switzerland, 2016. 2. Chui, K.T.; Alhalabi, W.; Pang, S.S.H.; Pablos, P.O.D.; Liu, R.W.; Zhao, M. Disease Diagnosis in Smart Healthcare: Innovation, Technologies and Applications. Sustainability 2017, 9, 2309. [CrossRef] 3. Spruit, M.; Lytras, M. Applied Data Science in Patient-centric Healthcare. Telemat. Inform. 2018, 35, 643–653. [CrossRef] 4. Khaire, N.N.; Joshi, Y.V. Disgnosis of disease using wrist pulse signal for classiﬁcation of pre-meal and post-meal samples. In Proceedings of the 2015 International Conference on Industrial Instrumentation and Control, Maharashtra, India, 28–30 May 2015; IEEE: Piscataway, NJ, USA. 5. Reddy, R.K.; Pooni, R.; Zaharieva, D.P.; Senf, B.; El Youssef, J.; Dassau, E.; Castle, J.R. Accuracy of Wrist-Worn Activity Monitors during Common Daily Physical Activities and Types of Structured Exercise: Evaluation Study. JMIR mHealth uHealth 2018, 6, e10338. [CrossRef] [PubMed] 6. Li, J.; Zhang, B.; Lu, G.; You, J.; Zhang, D. Body surface feature-based multi-modal Learning for Diabetes Mellitus detection. Inf. Sci. 2019, 472, 1–14. [CrossRef] 7. He, D.; Wang, L.; Fan, X.; Yao, Y.; Geng, N.; Sun, Y.; Xu, L.; Qian, W. A new mathematical model of wrist pulse waveforms characterizes patients with cardiovascular disease—A pilot study. Med. Eng. Phys. 2017, 48, 142–149. [CrossRef] [PubMed] 8. Qiao, L.J.; Qi, Z.; Tu, L.P.; Zhang, Y.H.; Zhu, L.P.; Xu, J.T.; Zhang, Z.F. The Association of Radial Artery Pulse Wave Variables with the Pulse Wave Velocity and Echocardiographic Parameters in Hypertension. Evid. Based Complement. Altern. Med. 2018, 2018, 5291759. [CrossRef] 9. Zhang, Z.; Zhang, Y.; Yao, L.; Song, H.; Kos, A. A sensor-based wrist pulse signal processing and lung cancer recognition. J. Biomed. Inform. 2018, 79, 107–116. [CrossRef] [PubMed] 10. Chen, Y.; Zhang, L.; Zhang, D.; Zhang, D. Computerized wrist pulse signal diagnosis using modiﬁed auto-regressive models. J. Med. Syst. 2011, 35, 321–328. [CrossRef] Appl. Sci. 2019, 9, 2284 14 of 14 11. Chow, W.H.; Wu, C.K.; Tsang, K.F.; Li, B.Y.S.; Chui, K.T. Wrist pulse signal classiﬁcation for inﬂammation of appendix, pancreas, and duodenum. In Proceedings of the 40th Annual Conference of the IEEE Industrial Electronics Society, Dallas, TX, USA, 30 October–1 November 2014; IEEE: Piscataway, NJ, USA. 12. Garg, N.; Bisht, A.; Ryait, H.S.; Kumar, A. Identiﬁcation of motion outliers in wrist pulse signal. Comput. Electr. Eng. 2018, 67, 776–790. [CrossRef] 13. Liu, X.; Ji, Z.; Tang, Y. Recognition of pulse wave feature points and non-invasive blood pressure measurement. J. Signal Process. Syst. 2017, 87, 241–248. [CrossRef] 14. Wang, D.; Zhang, D.; Lu, G. Generalized Feature Extraction for Wrist Pulse Analysis: From 1-D Time Series to 2-D Matrix. IEEE J. Biomed. Health Inform. 2017, 21, 978–985. [CrossRef] [PubMed] 15. WHO. Global Health Estimates 2015: Deaths by Cause, Age, Sex, by Country and by Region, 2000–2015; World Health Organization: Geneva, Switzerland, 2016. 16. Chui, K.T.; Tsang, K.F.; Chi, H.R.; Ling, B.W.K.; Wu, C.K. An accurate ECG based transportation safety drowsiness detection scheme. IEEE Trans. Ind. Inform. 2016, 12, 1438–1452. [CrossRef] 17. Montazeri, A.; West, C.; Monk, S.D.; Taylor, C.J. Dynamic modelling and parameter estimation of a hydraulic robot manipulator using a multi-objective genetic algorithm. Int. J. Control 2017, 90, 661–683. [CrossRef] 18. Tseng, F.H.; Wang, X.; Chou, L.D.; Chao, H.C.; Leung, V.C. Dynamic Resource Prediction and Allocation for Cloud Data Center Using the Multiobjective Genetic Algorithm. IEEE Syst. J. 2018, 12, 1688–1699. [CrossRef] 19. Chui, K.T.; Tsang, K.F.; Wu, C.K.; Hung, F.H.; Chi, H.R.; Chung, H.S.H.; Man, K.F.; Ko, K.T. Cardiovascular diseases identiﬁcation using electrocardiogram health identiﬁer based on multiple criteria decision making. Expert Syst. Appl. 2015, 42, 5684–5695. [CrossRef] 20. Fuchida, M.; Pathmakumar, T.; Mohan, R.E.; Tan, N.; Nakamura, A. Vision-based perception and classiﬁcation of mosquitoes using support vector machine. Appl. Sci. 2017, 7, 51. [CrossRef] 21. Wu, J.L.; Chang, P.C.; Tsao, C.C.; Fan, C.Y. A patent quality analysis and classiﬁcation system using self-organizing maps with support vector machine. Appl. Soft Comput. 2016, 41, 305–316. [CrossRef] 22. Ryan, W.L. Digestive Diseases—Research and Clinical Developments: Appendicitis: Symptoms, Diagnosis, and Treatments; Nova Science: New York, NY, USA, 2011. 23. Keyzer, C.; Gevenois, P.A. Imaging of Acute Appendicitis in Adults and Children; Springer: Berlin, Germany, 24. Serra, S.; Jani, P.A. An approach to duodenal biopsies. J. Clin. Pathol. 2006, 59, 1133–1150. [CrossRef] 25. Adams, D.B.; Cotton, P.B.; Zyromski, N.J.; Windsor, J. Pancreatitis: Medical and Surgical Management; Wiley Blackwell: Chichester, UK, 2017. 26. Yadav, D.; Timmons, L.; Benson, J.T.; Dierkhising, R.A.; Chari, S.T. Incidence, prevalence, and survival of chronic pancreatitis: A population-based study. Am. J. Gastroenterol. 2011, 106, 2192. [CrossRef] 27. De Haan, R.R.; Visser, J.B.; Pons, E.; Feelders, R.A.; Kaymak, U.; Hunink, M.M.; Visser, J.J. Patient-speciﬁc workup of adrenal incidentalomas. Eur. J. Radiol. Open 2017, 4, 108–114. [CrossRef] 28. Roberts, D.R.; Bahn, V.; Ciuti, S.; Boyce, M.S.; Elith, J.; Guillera-Arroita, G.; Hauenstein, S.; Lahoz-Monfort, J.J.; Schroder, B.S.; Thuiller, W.; et al. Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure. Ecography 2017, 40, 913–929. [CrossRef] 29. Deb, K. Multi-Objective Optimization Using Evolutionary Algorithms; John Wiley & Sons, Inc.: New York, NY, USA, 2001. 30. Herbrich, R. Learning Kernel Classiﬁers Theory and Algorithms; The MIT Press: London, UK, 2002. 31. Smits, G.F.; Jordan, E.M. Improved SVM regression using mixtures of kernels. In Proceedings of the 2002 International Joint Conference on Neural Networks, Honolulu, HI, USA, 12–17 May 2002; IEEE: Piscataway, NJ, USA. 32. Rangaprakash, D.; Dutt, D.N. Study of wrist pulse signals using time domain spatial features. Comput. Electr. Eng. 2015, 45, 100–107. [CrossRef] © 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

Journal

Applied Sciences – Multidisciplinary Digital Publishing Institute

Published: Jun 3, 2019

Keywords: bioinformations; genetic algorithm; multiobjective optimization; organ inflammation; support vector machine; wrist pulse signal

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

A Novel MOGA-SVM Multinomial Classification for Organ Inflammation Detection

A Novel MOGA-SVM Multinomial Classification for Organ Inflammation Detection

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

A Novel MOGA-SVM Multinomial Classification for Organ Inflammation Detection

A Novel MOGA-SVM Multinomial Classification for Organ Inflammation Detection

References (33)

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies