You are on page 1of 4

2017 IEEE Trustcom/BigDataSE/ICESS

Mapping ECG Signal Sequences on Variant Maps


Lihua Leng Jeffery Zheng
School of Software School of Software
Yunnan University Yunnan University
Kunming, China Kunming, China
lihualeng77@163.com conjugatelogic@yahoo.com

Abstract—Batch ECG signal sequences provide real-time, multiple statistical probability, the variant maps can be used to
dynamic measurements for groups of heart disease patients as process ECG signals from a global viewpoint.
special biomedical big datasets. To extract various diseases
information, we propose a new method using multiple statistical The ECG datasets used in this paper provided by the First
probability distributions to process batch ECG signal sequences People's Hospital of Yunnan Province, all the ECG datasets
on variant maps. Using this approach, it is more convenient to have diagnosed information by the doctors. The whole datasets
identify normal ECG datasets and abnormal ECG datasets on are composed of two parts: normal ECG datasets and abnormal
distinct distributions. Sample cases are illustrated. ECG datasets in convenient comparison. Both normal ECG and
abnormal ECG datasets are used to select suitable sample files
Keywords—ECG signal sequences; Biomedical big datasets;
Variant map in testing.

I. INTRODUCTION II. TRANSFORMATION


Electrocardiography (ECG) is the process of recording the In this ECG dataset, there are 136300 abnormal cases and
electrical activity of the heart over a period of time using 133178 normal cases. Each case contains heart-rate wave, p
electrodes placed on the skin. These electrodes detect the tiny wave, qr wave and other 14 records and the relevant diagnostic
electrical changes on the skin that arise from the heart muscle's result. The initial step of test is selecting the same number of
electric-physiologic pattern of depolarizing and repolarizing records from normal and abnormal ECG. The main process
during each heart beat [1]. Cardiac arrhythmia is a group of consists of four steps: Preprocess, Transform, Meta measures
conditions in which the heartbeat is irregular. It is a very and Combinatorial maps from original ECG signal sequences
commonly performed cardiology test. ECG's are normally to variant maps.
printed on a grid in the form of electrocardiogram [2], the Step 1: Preprocess
horizontal axis represents time and the vertical axis represents In this step, proper values in data files are checked.
voltage. Interpretation of the ECG is fundamentally about Including exception value exclusion, exception value
understanding the electrical conduction system of the heart. interpolation and datasets normalization. For example, the
Normal conduction starts and propagates in a predictable range of normal p wave is between -179:178, if a value of 5000
pattern, and deviation from this pattern can be a normal is present, it is an exception value.
variation or be pathological for the heart either normal or
abnormal. Interpretation of the ECG is ultimately that of Step 2: Transform
pattern recognition.
From original sequence to quaternions
The research of the ECG signal has an important practical A multiple valued ECG sequence will be transformed into a
significance. Associated with advanced machine learning quaternion sequence on four meta states {+, −, ⊤, ⊥} which
technologies in past 50 years, the automated ECG resemble the quaternions {Plus, Minus, Top, Bottom}. Let X
interpretation has being a useful tool when access to a local difference, L local average, T global difference, V global
specialist is not possible. Although considerable effort has been average, A i-th sample and A next neighbor. For an ECG
made to improve automated ECG algorithms, the sensitivity of sequence with N elements, four parameters X, T, L and V are
the automated ECG interpretation is of limited value and it is formulated in the equations (1)-(4).
always difficult to handle huge amount of monitoring ECG
sequences in big datasets for many heart diseases. X= − (1)
∑ | |
From an analysis viewpoint, ECG sequence is a special T= (2)
type of pseudo random dataset. The research of ECG signals is L= (3)
a hot topic in bioinformatics academia. Scholars in fields

research on the randomness of ECG sequences: ECG signal V= (4)
compression [3], acquisition [4], de-noising [5] wavelet
transform [6] and neural network [7] and statistical probability Applying the four parameters, the original ECG sequence
approaches. Due to the simplicity of its process, applying A . . . A is transformed into a sequence of quaternions B . . . B
by using formula (5):

National Science Foundation of China NSFC (No. 61362014)


Overseas Higher-level Scholar Project of Yunnan Province, China (No. W8110305).

2324-9013/17 $31.00 © 2017 IEEE 881


DOI 10.1109/Trustcom/BigDataSE/ICESS.2017.326
> > 0: =+ Under the nine segment lengths shown in Fig. 2 and Fig. 3,
> ≤ 0: =− the abnormal cases are more tight than the normal cases.
(5) Associated increasing the length of segment, whole map will
≤ > : =⊤
≤ ≤ : =⊥ be grouped to a smaller region.
As an example, the input ECG dataset is {1,5,3,2,0,9,20,10}
V. CONCLUSION
, the X={4,-2,-1,-2,9,11,-10}, the values T=1.28, the values
L={3,4,2.5,1,4.5,14.5,15} and the values V=6.35. Based on It is an initial exploration to project whole ECG signal
these values, we can finally obtain the output sequence sequence into a 2D variant map. From selected cases, there
{+, ⊥, ⊥, ⊥, +, +, ⊤}, are significant differences among normal and abnormal
cases with refined details. Further explorations on larger
Step 3: Meta measures datasets and more refined cases are required.
Getting the quaternion sequence . . . in step 2, it will
be divided into multiple segments of length M in a total number VI. ACKNOWLEDGMENT
of S segments ( S = N/M ). For each segment, four meta- Thank to the First People's Hospital of Yunnan Province for
measures { , , ⊺ , } are calculated, each measure ECG datasets, to National Science Foundation of China NSFC
represents one of quaternions in the segment. Four measures (No. 61362014) and the Overseas Higher-level Scholar Project
meet 0 ≤ , , ⊺, ≤ M, and + + ⊺+ = of Yunnan Province, China (No. W8110305) for financial
M. After processing all S segments, the quaternion sequence supports to the project.
becomes a set of four meta measures as a meta measure set VII. REFERENCES
with S tuples.
[1] Mcsharry P E, Clifford G D, Tarassenko L, et al. A dynamical model for
As the example shown in step 2, the output sequence is generating synthetic electrocardiogram signals.[J]. IEEE Trans Biomed
{+, ⊥, ⊥, ⊥, +, +, ⊤}, so four meta-measures are = 3, = Eng, 2003, 50(3):289-94.
0, ⊺ = 1, = 3. [2] Xinyan F U, Huang L, Jiang W, et al. The application of
electrocardiogram and vectorcardiogram experiment operation board in
Step 4: Combinatorial maps teaching[J]. Zhejiang Medical Education, 2015.
Using the meta measure set obtained in step 3, there are [3] Kim H, Yazicioglu R F, Merken P, et al. ECG signal compression and
many combinatorial schemes to use four measures to project classification algorithm with quad level vector for ECG holter system[J].
IEEE Transactions on Information Technology in Biomedicine, 2010,
these measures into 2D density scatter plots as variant maps. In 14(1):93-100.
this paper, a special projection is selected to determine a pair of [4] Gupta R, Bera J N, Mitra M. Development of an embedded system and
(X, Y) for an axis position: two items selected for X and three MATLAB-based GUI for online acquisition and analysis of ECG
items selected for Y. For each meta measures, X = + ⊺, signal[J]. Measurement, 2010, 43(9):1119-1126.
and Y = + ⊺+ . Collecting all segments clustering on [5] Sharma L N, Dandapat S, Mahanta A. ECG signal denoising using
higher order statistics in Wavelet subbands[J]. Biomedical Signal
possible (X, Y) positions, their Z values are obtained as a 2D Processing & Control, 2010, 5(3):214-222.
map. [6] And A Q, Daqrouq K. ECG Signal Enhancement Using Wavelet
Transform[J]. Wseas Transactions on Biology & Biomedicine, 2010,
In convenient to understand the transformation process in 7(2).
the step 4, a visual example is shown in Fig. 1. (a) p wave of [7] Rai H M, Trivedi A, Shukla S. ECG signal processing for abnormalities
abnormal ECG dataset is selected on N = 136300, M = 66. A detection using multi-resolution wavelet transform and Artificial Neural
white circle is at the position X = 35 and Y = 52, its color Network classifier[J]. Measurement, 2013, 46(9):3238-3246.
indicates Z = 10 . i.e. there are 10 segments to meet + [8] Zheng Z, Maeder AJ (1992) The conjugate classification of the kernel
form of the hexagonal grid, Springer Japan, Tokyo.
⊺ = 35 and + ⊺+ = 52. The white areas indicate
[9] Zheng ZJ (1994) Conjugate transformation of regular plan lattices for
no projection point on this sequence. binary images, PhD thesis, Monash University.
[10] Zheng J, Zhang W, Luo J, Zhou W, Shen R (2013) Variant map system
III. TESTING RESULTS to simulate complex properties of DNA interactions using binary
In convenient in comparison, two types of ECG signals are sequences. Advances in Pure Mathematics 3: 5.
selected. [11] Qinxian Bu,Jeffrey Z.J.Zheng.2D Conjugate Maps of DNA
Sequences[J].Journal of Information Security,2013,4,193-196.
In Fig. 1, both abnormal and normal cases are selected on [12] Zheng J, Luo J, Zhou J (2014) Pseudo DNA sequence generation of non-
M=66. In Fig. 2, nine abnormal cases are selected on coding distributions using variant maps on cellular automata. Applied
M={99,88,77,66,55,44,33,22,11}. In Fig. 3, nine normal cases Mathematics 5: 153.
are selected on M={99,88,77,66,55,44,33,22,11}. [13] Zheng J, Zhang W, Luo J, Zhou W, Liesaputra V (2014) Variant map
construction to detect symmetric properties of genomes on 2D
distributions. Journal of Data Mining in Genomics & Proteomics 5: 1.
IV. COMPARISON
[14] Heim DM, Heim O, Zeng PA, Zheng J (2016) Successful Creation of
There are significant differences to compare two maps in Regular Patterns in Variant Maps from Bat Echolocation Calls. Biol Syst
Fig. 1. The 2D map of the abnormal case is clustered in a small Open Access 5: 166. doi:10.4172/2329-6577.1000166.
region and the 2D map of the normal case has a wider [15] Zheng J, Zheng C (2011) Variant measures and visualized statistical
distributions. Acta Photonica Sinica 40: 1397.
distribution.
[16] Zheng J, Zheng C (2012) Variant simulation system using quaternion
structures. Journal of Modern Optics 59: 484.

882
(a) (b)

Figure 1. Two types of ECG sequence in the variant maps in M = 66. Fig.(a) is abnormal ECG sequence; Fig.(b) is normal ECG sequence

883
Figure 2. Nine cases of abnormal ECG sequence in N = 136300, M = {99,88,77,66,55,44,33,22,11}

Figure 3. Nine cases of normal ECG sequence in N = 133178, M = {99,88,77,66,55,44,33,22,11}

884

You might also like