Linear and non-linear analysis of cardiac health in diabetic subjects
Introduction
Diabetes mellitus, or diabetes, is a chronic disease, which is characterized by hyperglycaemia. Hyperglycaemia, is a metabolic disorder, were excess glucose is present in the blood. This results in an elevated blood glucose level, which leads to serious detrimental consequences. The disease affects eye (retinopathy) [1], [2], nerves (neuropathy) [3], kidney (nephropathy) [4] and heart (cardiomyopathy) [5].
According to World Health Organization (WHO) more than 220 million people worldwide had diabetes, in 2009. It is estimated that this figure will increase to 440 million by the year 2030 [6]. In 2008, the American Diabetes Association reported 23.6 million (approximately 7.8% of the population) children and adults in the United States have diabetes [7].
Cardiovascular disease (CVDs) is the number one cause of death globally. WHO estimated about 29% of deaths were due to CVDs (totaling 17.1 million) in 2004. They projected about 23.6 million people will succumb to the disease by 2030 [8].
Heart rate variation (HRV) is the name of a biological time series signal which indicates the variation of heart rate between two consecutive heart beats [9]. HRV is a non-invasive tool to assess the autonomic nervous system (ANS). HRV may take precedence over the situation where loads and loads of data are to be collected for several hours in order to understand and identify abnormalities. Thus, HRV can be seen as highly effective diagnostic tool. In recent years, there has been much work by various researchers on the analysis of heart rate variability [10], [11], [12]. HRV also gives information about the sympathetic–parasympathetic autonomic balance and about the risk of sudden cardiac death in these patients [13]. HRV measurements are easy to obtain and they are reproducible, if measured under standardized conditions [14], [15].
Biological time series analysis can be done in time and frequency domain as well as with non-linear methods. The aim of the analysis is to detect important dynamical properties of the physiological phenomena which are hidden in the data. A particular problem of biological time series analysis comes from the fact that statistical characteristics can vary with time. This makes time domain analysis unreliable. Frequency domain parameters give better assessment of the autonomic function, but the reliability of spectral power diminishes with the decrease in power signal and signal-to-noise ratio [16].
Non-linear dynamical techniques are used in many areas including biology and medicine, because they can overcome the shortcomings of time and frequency domain methods [17]. These techniques yield useful indicators of pathologies, because many biological systems, such as the cardiovascular system, are complex and can never be linear in nature. Schumacher [18] have explained the use of linear and non-linear techniques in the analysis of HR signals. Furthermore, non-linear methods have been applied in tracking HRV signals and predicting the onset of dangerous cardiovascular related events, such as Ventricular Tachycardia and congestive heart failure [19].
The decreased beat-to-beat variability during deep breathing of patients with diabetic neuropathy was first reported by Wheeler and Watkins [20] and confirmed by many others. Studies which compared cardiac autonomic function tests and HRV indices (based on both short (5-min) and 24-h electrocardiogram (ECG) recordings), showed that in diabetic patients without abnormal function tests HRV was lowered [21]. It was concluded that cardiac (parasympathetic) autonomic activity was diminished in diabetic patients before clinical symptoms of neuropathy became evident [22], [23].
In this paper we analyze HRV signals with both linear and non-linear methods. The non-linear methods uncover subtle changes in the beat to beat variation of the signal. This makes them more robust against systematic (linear) measurement errors. Furthermore, these methods aim to reveal the underlying physiological processes directly and not, as in linear methods, measure secondary or indirect signal phenomena. Therefore, we promote non-linear methods to determine the cardiac health of diabetic patients. These measures can be used for treatment monitoring and to assess the results of clinical trials.
Section snippets
Data
The electrocardiograms (ECG) of 15 patients (10 male and 5 female) with diabetes and of 15 healthy volunteers (8 male and 7 female)) were recorded, with the patients in a relaxed supine position for 60 min. The subjects under study of diabetes were in the age group of 50 ± 70 years (mean ± standard deviation = 58.5 ± 6.42 years) and the duration of diabetes for the patient groups was 5 ± 15 years. The normal subjects were in the age group of 40 ± 60 years (mean ± standard deviation = 50 ± 8.8 years). ECG signals
Time domain
Table 1 shows t-test analysis [46] for the time domain parameters. All features are statistically significant as they possess p-values of less than 0.0001. Even though time domain parameters, such as , NN50 and PNN50, are widely used, it can be shown that results can be affected by disturbances like artifacts and noise. To effectively measure these parameters, sufficient measures have to be taken to eliminate them prior to data collection. This careful measurement is reflected in the low p
Discussion
We have shown the results of time domain, frequency domain, and non-linear analysis in the previous section. We have used reduced p-values to analyze both nonlinear and linear feature extraction methods. From past experience we know that features with a low p-value are more useful when it comes to differentiating normal and type 2 diabetic individuals. However, strictly speaking, a lower p-value just indicates that the relationship between the groups in not linear. Therefore, all we can claim
Conclusion
Diabetes, a metabolic disorder, is one of the most dangerous diseases, even though, like HIV/AIDS, not a direct killer, it seriously shortens life. HRV measurements are used as base signals to assess the cardiac health of diabetic and normal subjects.
In this report, we have extracted six non-linear features from heart rate signals from both normal and diabetes patients. The features are as follows: SD2, REC, DET, ApEn, SampEn and CD. Both ApEn and SampEn show lower entropy values for the
Acknowledgements
Authors thank Ms Loh Khei Ying for helping in running the non-linear codes on normal and diabetes heart rate signals.
References (56)
- et al.
The eyes in diabetes and diabetes through the eyes
Diabetes Research and Clinical Practice
(2007) - et al.
Stability over time of variables measuring heart rate variability in normal subjects
The American Journal of Cardiology
(1991) - et al.
Association of hyperglycemia with reduced heart rate variability (the framingham heart study)
The American Journal of Cardiology
(2000) - et al.
Risk stratification for arrhythmic events in postinfarction patients based on heart rate variability, ambulatory electrocardiographic variables and the signal-averaged electrocardiogram
Journal of the American College of Cardiology
(1991) - et al.
Influence of high-frequency bandwidth on heart rate variability analysis during physical exercise
Biomedical Signal Processing and Control
(2007) - et al.
Patterns of beat-to-beat heart rate variability in advanced heart failure
American Heart Journal
(1992) - et al.
Measuring the strangeness of strange attractors
Physica D: Nonlinear Phenomena
(1983) - et al.
Testing for nonlinearity in time series: the method of surrogate data
Physica D: Nonlinear Phenomena
(1992) - et al.
A dynamic model to characterize beat-to-beat adaptation of repolarization to heart rate changes
Biomedical Signal Processing and Control
(2008) - et al.
Heart rate changes in diabetes mellitus
The Lancet
(1981)
Spectral analysis of heart rate fluctuations. A non-invasive, sensitive method for the early diagnosis of autonomic neuropathy in diabetes mellitus
Journal of the Autonomic Nervous System
Automatic identification of cardiac health using modeling techniques: a comparative study
Information Science
Novel approach for fetal heart rate classification introducing grammatical evolution
Biomedical Signal Processing and Control
Predictability of baroreflex sensitivity induced by phenylephrine injection via frequency domain indices computed from heart rate and systolic blood pressure signals during deep breathing
Biomedical Signal Processing and Control
Algorithms for the automated detection of diabetic retinopathy using digital fundus images: a review
Journal of Medical Systems
Diagnosis and management of diabetic autonomic neuropathy
British Medical Journal
The Diabetes Sourcebook
Diabetes and cardiovascular disease: a statement for healthcare professionals from the American Heart Association
Circulation
Comprehensive analysis of cardiac health using heart rate signals
Physiological Measurement
Heart rate variability in relation to prognosis after myocardial infarction: selection of optimal processing techniques
European Heart Journal
Heart rate variability: standards of measurement, physiological interpretation, and clinical use
Circulation
Heart rate variability: origins, methods, and interpretive caveats
Psychophysiology
Chaos theory, heart rate variability, and arrhythmic mortality
Circulation
Cardiac arrhythmia classification using autoregressive modeling
BioMedical Engineering OnLine
Heart rate variability: a review
Medical and Biological Engineering and Computing
Cited by (65)
In-silico cardiovascular hemodynamic model to simulate the effect of physical exercise
2023, Biomedical Signal Processing and ControlA comparative analysis on diagnosis of diabetes mellitus using different approaches – A survey
2020, Informatics in Medicine UnlockedCitation Excerpt :As linear methods are unable to find hidden information in HR signals. Researchers in [40] showed that clinically significant nonlinear parameters were correlation dimension approximate entropy, sample entropy, and recurrence plot properties. The AdaBoost classifier is considered as the best classifier for diabetes diagnosis using HR signals.
Automated detection of diabetic subject using pre-trained 2D-CNN models with frequency spectrum images extracted from heart rate signals
2019, Computers in Biology and MedicineCitation Excerpt :The same group [20] developed a novel diabetes index approach for diagnosis of diabetic neuropathy using features extracted from HRV signals. In Ref. [21], time, frequency, and nonlinear domain techniques were utilized to analyze normal and diabetic HR signals. They showed that nonlinear HRV analysis is more effective than time and frequency methods.
Cardiac, diabetic and normal subjects classification using decision tree and result confirmation through orthostatic stress index
2019, Informatics in Medicine UnlockedThe Use of Empirical Mode Decomposition on Heart Rate Variability Signals to Assess Autonomic Neuropathy Progression in Type 2 Diabetes
2023, Applied Sciences (Switzerland)