International Journal of Pure and Applied Mathematics Volume 117 No. 10 2017, 87-90


ISSN: 1311-8080 (printed version); ISSN: 1314-3395 (on-line version)


doi: 10.12732/ijpam.v117i10.18



S.Vyshali1, Dr.M.V.Subramanyam2, Dr.K.Soundara Raajan3

Assistant Professor, ECE Department, G.Pulla Reddy Engg.College, Kurnool.,AP, India

Principal, Santhi Ram, Nandyal. Kurnool,A.P,India.

Retd.Professor, JNTU, Ananthapuramu. A.P,India.

[email protected],[email protected] , [email protected]


Abstract: In the diagnosis of MRI signal, Atrial fibrillation is a common disease observed. In the coding of MRI analysis, regressive models were used to detect the MRI features for atrial fibrillation detection. the regressive model is used as feature extraction process, where, the MRI signal are processed for a regular time interval to extract the characteristic variation in MRI signal and develop a classification based on the variations observed in feature detection. The regression model are however time consuming due to recurrent coding and does not focus on the feature divergence issue.

As the feature extracted for different test subjects for the same observations varies, the divergence issue need to be addressed to achieve a optimal solution in detection. in this paper divergence problem using Bergman divergence approach is developed to achieve optimal recognition observations in MRI analysis.

Keywords: atrial fibrillation; auto regression model; classification performance; divergence optimization; mri analysis.

1. Introduction

With the emergence of new technologies the real time world is tending towards automation processing. Various technologies have emerged in recent past to achieve higher processing efficiency to improve the system performance and higher accuracy.

Get quality help now

Proficient in: Biology

5 (339)

“ KarrieWrites did such a phenomenal job on this assignment! He completed it prior to its deadline and was thorough and informative. ”

+84 relevant experts are online
Hire writer

In various such applications the process of automation is also tending towards medical diagnosis. In the application of medical diagnosis the captured samples are passed to the processing unit and advance computational algorithms are executed to achieve the finest results. In various such applications, MRI processing in automated diagnosis is emerging. During the process of compression or during the capturing process there are artifacts generated which would results in wrong interpretation. In the identification of MRI signal, Atrial fibrillation is the usual cardiac arrhythmia among people, by growing commonness in the old people[1]. During the period following a surgical operation, after thoracic surgery, it is also one of the most recurrent method [2]. Thus donating to extension of hospitalization and to an enlarge of the related price[3]. Although declining years alone appear to be the powerful soothsayer for the evolution of medicinal in the last

decennary [4]. Many studies have focused on finding computation able to forecast medicinal by the examination of surface electrocardiographic reports [5]. A risk hierarchal based on surgical tests would be very practical either to modify the precaution anti-arrhythmia treatment such as narcotic (or) in sick people prone to grow surgical medicinal shortening patient hurt and lowering price associated to the hospitalization (or) to bound narcotic managing in low risk subjects[6]. Atrial remodeling is defined as any change in atrial structure or function that promotes atrial arrhythmias is central to AF[7].According to theory proposed in [8,9], fastly launch atrial ectopic foci can result in early atrial multiplex in episodes of atrial cardiac arrest for the initiation of the arrhythmia on a make prone abnormal substrate. In [10] the classification of MRI signal to detect atrial fibrillation was suggested. The Approach developed a auto regressive model to achieve the objective of feature extraction. The coefficients are measured for regular time interval extracted through Burg’s method. A KNN classifier is used to detect the medical over different data length and classification is made over MIT-BIH medical database. To present the develop work, the rest of the paper is presented in 5 sections, where a basic detail to MRI signal and its analysis for medical is presented. The conventional modeling of auto regression model is suggested in section 3. Section 4 present the proposed approach of weighted divergence function, the experimental results are outlined in section 5. Section 6 present the conclusion of the developed work.

2. Autoregressive Modeling [14]

For the analysis of Atrial fibrillation, the MIT-database is used. The database consist of 23 records of medical signals which are sampled as 250Hz, and 18 Normal sinus rhythms. To process the feature diagnosis a auto regressive model is used. The procedure calculates the characteristic as a set of 1-D forecast model, where the regression model, directly think about as at the same time congregate of more properties of the file. Auto-regression is a stochastic process used in statistical calculations in which future values are estimated based on a weighted sum of past values. In common, Auto-regression detects the subsets of ascribes and their things at the same time of a data matrix using



International Journal of Pure and Applied MathematicsSpecial Issue

a particular basis. Different from congregate methods, Auto-regression also explains a congregate basis and then modify it. Auto-regression gives notable profits. Basically, two types of Auto-regression approaches are there. They are Block Auto-regression and Information Theoretic Auto-regression(ITAR). Among the two, the ITAR proposal was demonstrated to be ideal. The ITAR was founded on Burg’s method.

For a given Auto- regression model with (R,C) and a matrix approximation scheme P, a class of random variables which store the properties of attribute data matrix X is defined. The objective function tries to reduce the attributes loss on the approximation of ? for a Auto- regression R, C. The burg’s prediction information of X defined by [14]:

( )?( )( ()


expected value and the divergence for an ideal congregate as follows

()(? )

Here, the Auto- regression was straightly connected with Euclidean distance and can be shown as


Here, the Auto- regression was estimated by just manipulating the Euclidean distance between the ascribes expanse values.

3. Weighted-ITAR Approach

In this proposed method, a weighted ITAR coding for co-clustering of data type is presented. In the approach of co-clustering operation following ITAR, the clustering are made based on the divergence factor of two observations(x1,x2).The approach of co-clustering using weight factor is defined by allocating a weight value to the distinct class attributes in the sub cluster.

Where the class attributes is defined as in table 1. The dataset is clustered into sub clusters based on the criterion of Bergman divergence Information ( ), satisfying the convergence problem,

((? )


Whereand? is the new co-cluster formed for the sub

cluster. Here,is defined as the divergence operator given

by the Euclidian distance of the two observing element (),

This section illustrates the details about the result analysis. The performance of proposed approach was verified through accuracy. The accuracy of the proposed approach was measured


In consideration to the stated approach, the weighted co-clustering computes an aggregative weight value of each of the class label ( ) given as,


Table 1: Class label attribute to distinct classes formed


For this set of dataset, to apply weighted co-clustering process


the following step of operations are applied,

Step 1: The Bergman divergence for the two observing element is computed using (6).

Step 2: For each divergence, ( ), the convergence

criterion defined by (5) is computed. (2) Step 3: The aggregative sum weighted value for the observing class is then computed using (7).

Step 4: The convergence criterion is then checked for the obtained aggregated weight value following (9).

Step 5: The above steps are repeated for all sub classes to co-

cluster. (3)

In the considered case, the total weight

While co-clustering the elements of the entire sub class is obtained by,

= 15.

Now, for the clustering of two sub classes,and, the

aggregative weight sum is obtained as,


Considering convergence problem of (9), the limiting value is 15/5 = 3.

In this casehence not clustered in the same group,


even though the divergence criterion is satisfied. This results in co-clustering of sub-classes with highest similarity into one class. As, could be observed, class-1 and class-4 are of more dissimilarity than class 1and 2. The allocated weight values governed the class relation and the limiting boundness of the total weight value prevents the wrong clustering of the sub-class. The operational algorithm for the suggested weighted co-clustering approach is as illustrated below.

4. Experimental results

using the standard confusion metrics. The metrics are listed as True Positive (TP), False Positive (FP), True Negative (TN), False negative (FN). The complete analysis was carried out on



International Journal of Pure and Applied MathematicsSpecial Issue

all the classes of patients listed above such as Normal (0), Effective 1 (1), Effective 2 (2), Effective 3 (3) and Effective 4

(4). Initially, the TP denotes the Effective people correctly identified as effective, FP denotes the Normal people incorrectly identified as effective, TN denotes Normal people correctly

Table.2. Accuracy Comparison for proposed approach with earlier approaches under all classes


Table.3. performance table for the developed approach over conventional model for 5 test observation


Table.4 performance table for the developed approach over conventional model for 15 test observation


Table.5. performance table for the developed approach over conventional model for 30 test observation


5. Conclusion

A current advance of attribute description for medicinal detection based on the comparative belongings of the data ascribes and kind evaluate is shown. In the extract procedure for heart disease study, the congregate operation based on data type things of constant and separate is used. The sub-cluster emergence results in ideally data array with medicinal

identified as Normal and finally FN denotes Effective people incorrectly identified as Normal. The accuracy at each and every stage was evaluated by the following mathematical expression,


characteristics choice in discussion with its class belongings. The class belongings is shown to be a notable examination in suitable clustering of dataset and shown a more presentation in categorization perfection.

6. References

Chih-Chun Chia, James Blum, Zahi Karam, Satinder Singh,

Zeeshan Syed, “Predicting Postoperative Atrial Fibrillation from Independent MRI Components”, Association for the

Advancement of Artificial Intelligence, 2014.

L. Clavier, J. Boucher, R. Lepage, J. Blanc, and J. Cornily, “Automatic p-wave analysis of patients prone to atrial fibrillation,” Medical and Biological Engineering and

Computing, vol. 40, pp. 63–71, 2002.

E. Ros, S. Mota, F. Fernandez, F. Toro, and J. Bernier, “MRI characterization of paroxysmal atrial fibrillation: Parameter extraction and automatic diagnosis algorithm,” Computers in

Biology and Medicine, vol. 34, no. 8, pp. 679 – 696, 2004.

B. Pourbabaee and C. Lucas, “Automatic detection and prediction of paroxysmal atrial fibrillation based on analyzing

MRI signal feature classification methods,” in 2008 Cairo

International Biomedical Engineering Conference, CIBEC 2008, Cairo, Egypt, 2008.

B. Hickey, C. Heneghan, and P. De Chazal, “Non-episode-dependent assessment of paroxysmal atrial fibrillation through measurement of RR interval dynamics and atrial premature contractions,” Annals of Biomedical Engineering, vol. 32, no. 5, pp. 677 – 687, 2004.

D. Kim, Y. Seo, W. R. Jung, and C.-H. Youn, “Detection of long term variations of heart rate variability in normal sinus rhythm and atrial fibrillation MRI data,” in International

Conference on BioMedical Engineering and Informatics, BMEI 2008, vol. 2, Sanya, Hainan, China, 2008, pp. 404 – 408.

N. Lentz, N. Kikillus, and A. Bolz, “A screening method to detect atrial fibrillation with symbolic dynamics,” in IFMBE

Proceedings, vol. 25, no. 4, Munich, Germany, 2009, pp. 1464 – 1467.

Steinberg JS, Zelenkofske S, Wong SC, Gelernt M, Sciacca R, Menchavez E. Value of the P wave signal-averaged MRI for predicting atrial fibrillation after cardiac surgery. Circulation 1993;88:2618-22.

Stafford PJ, Kolvekar S., Cooper J. Signal averaged P wave compared with standard electrocardiography for prediction of atrial fibrillation after coronary artery bypass grafting. Heart 1997;77:417-22.

K.Padmavathia, K.Sri Ramakrishna, “Classification of MRI signal during Atrial Fibrillation using Autoregressive modeling”,

Procedia Computer Science 46 ( 2015 ) 53 – 59.




Cite this page

Divergence Coding. (2019, Dec 13). Retrieved from

Let’s chat?  We're online 24/7