Mapping ECG Signal Sequences on Variant Maps
Lihua Leng
School of Software
Yunnan University
Kunming, China
lihualeng77@163.com
Jeffery Zheng
School of Software
Yunnan University
Kunming, China
conjugatelogic@yahoo.com
Abstract—Batch ECG signal sequences provide real-time,
dynamic measurements for groups of heart disease patients as
special biomedical big datasets. To extract various diseases
information, we propose a new method using multiple statistical
probability distributions to process batch ECG signal sequences
on variant maps. Using this approach, it is more convenient to
identify normal ECG datasets and abnormal ECG datasets on
distinct distributions. Sample cases are illustrated.
Keywords—ECG signal sequences; Biomedical big datasets;
Variant map
I. INTRODUCTION
Electrocardiography ECG is the process of recording the
electrical activity of the heart over a period of time using
electrodes placed on the skin. These electrodes detect the tiny
electrical changes on the skin that arise from the heart muscle's
electric-physiologic pattern of depolarizing and repolarizing
during each heart beat [1]. Cardiac arrhythmia is a group of
conditions in which the heartbeat is irregular. It is a very
commonly performed cardiology test. ECG's are normally
printed on a grid in the form of electrocardiogram [2], the
horizontal axis represents time and the vertical axis represents
voltage. Interpretation of the ECG is fundamentally about
understanding the electrical conduction system of the heart.
Normal conduction starts and propagates in a predictable
pattern, and deviation from this pattern can be a normal
variation or be pathological for the heart either normal or
abnormal. Interpretation of the ECG is ultimately that of
pattern recognition.
The research of the ECG signal has an important practical
significance. Associated with advanced machine learning
technologies in past 50 years, the automated ECG
interpretation has being a useful tool when access to a
specialist is not possible. Although considerable effort has been
made to improve automated ECG algorithms, the sensitivity of
the automated ECG interpretation is of limited value and it is
always difficult to handle huge amount of monitoring ECG
sequences in big datasets for many heart diseases.
From an analysis viewpoint, ECG sequence is a special
type of pseudo random dataset. The research of ECG signals is
a hot topic in bioinformatics academia. Scholars in fields
research on the randomness of ECG sequences: ECG signal
compression [3], acquisition [4], de-noising [5] wavelet
transform [6] and neural network [7] and statistical probability
approaches.
Variant map is a useful scheme proposed in [8], then
expanded in [9]. It has being used in noncoding DNA sequence
detection [10-13], bats echolocation classification [14],
quantum interaction simulation [15,16]. Applying multiple
statistical probability, the variant maps can be used to process
ECG signals from a global viewpoint.
The ECG datasets used in this paper provided by the First
People's Hospital of Yunnan Province, all the ECG datasets
have diagnosed information by the doctors. The whole datasets
are composed of two parts: normal ECG datasets and abnormal
ECG datasets in convenient comparison. Both normal ECG and
abnormal ECG datasets are used to select suitable sample files
in testing.
II. TRANSFORMATION
In this ECG dataset, there are 136300 abnormal cases and
133178 normal cases. Each case contains heart-rate wave, p
wave, qr wave and other 14 records and the relevant diagnostic
result. The initial step of test is selecting the same number of
records from normal and abnormal ECG. The main process
consists of four steps: Preprocess, Transform, Meta measures
and Combinatorial maps from original ECG signal sequences
to variant maps.
Step 1: Preprocess
In this step, proper values in data files are checked.
Including exception value exclusion, exception value
interpolation and datasets normalization. For example, the
range of normal p wave is between -179:178, if a value of 5000
is present, it is an exception value.
Step 2: Transform
From original sequence to quaternions
A multiple valued ECG sequence will be transformed into a
quaternion sequence on four meta states which
resemble the quaternions {Plus, Minus, Top, Bottom}. Let
local difference, local average, global difference, global
average,
i-th sample and
next neighbor. For an ECG
sequence with elements, four parameters , , and are
formulated in the equations (1)-(4).
(1)
(2)
(3)
(4)
National Science Foundation of China NSFC (No. 61362014)
Overseas Higher-level Scholar Project of Yunnan Province, China (No. W8110305).