IIT Madras is a public technical and research university located in Chennai, Tamil Nadu. Founded in 1959, it is recognised as an Institute of National Importance.

IIT Madras has been ranked as the top engineering institute in India for four years in a row by the National Institutional Ranking Framework of the MHRD

It currently offers undergraduate, postgraduate and research degrees across 16 disciplines in Engineering, Sciences, Humanities and Management. About 596 faculty belonging to science and engineering departments and centres of the Institute are engaged in teaching, research and industrial consultancy.

IIT Madras

It is shown that the discrete cosine transform (DCT) of a given set of electrocardiographic (ECG) cycles can be approximated to a set of narrow band signals. These signals are estimated by performing singular value decomposition (SVD) on a prediction matrix made up of the DCT coefficients. The QRS pulses are then separated out in the DCT domain from a reduced rank approximation of the prediction matrix. The size of this reduced rank matrix is obtained by means of an optimally established orthogonality relationship between two complementary subspaces derived from the signal space of DCT coefficients.

Department of Applied Mechanics

World University root node

Signal Processing

A QRS estimator using linear prediction approach

The signal processing steps for the analysis of stress ECGs are aimed at improving the signal to noise ratio (SNR) of recordings in addition to eliminating artifacts due to respiration, movement of arms, etc. In this paper, we bring forth two important applications of the discrete cosine transform (DCT) for noise suppression and removal of baseline wander. The noise suppression algorithm has been framed on the basis of a two step procedure involving singular value decomposition (SVD) smoothing operation in transform domain followed by that in time domain. The mean square error (MSE) resulting from the first step is shown to effectively follow the trend obtained by using an ideal Wiener filter using DCT. In the second step, the degree of closeness to the minimum mean square error (MMSE) of the ideal Wiener filter is improved by subjecting the filtered outputs to a second SVD smoothing operation in time domain. Application of this scheme to noisy records has resulted in near perfect reproduction of the original noise free ECG without significant alterations in its morphological features.

GENERAL

Computers in Biology and Medicine

Data processing of stress ECGs using discrete cosine transform

An algorithm is presented for identification of premature ventricular cycles (PVCs) by performing a beat-by-beat analysis of the electrocardiogram (ECG). The circular convolution-multiplication relationships for the discrete cosine transform (DCT) similar to that for the discrete Fourier transform (DFT) have been used to decompose a windowed ECG cycle into system and excitatory functions representing action potential and excitation pattern of the heart muscle during cardiac cycle. The energy packing efficiency (EPE) derived from the DCT coefficients characterises the decay rate of the DCT of the system function and is related to bandwidths of resonances in the AR spectrum. EPE evaluated for the DCT of system response is shown to provide a sufficient criterion for the identification of PVCs. The algorithm was able to successfully identify PVCs from the recordings of an MIT-BIH database. Under noisy conditions, the algorithm clearly distinguishes PVC patterns from those of normal beats up to a signal-to-noise ratio (SNR) of 10 dB.

Applied Signal Processing

Identification of premature ventricular cycles in surface electrograms using the envelope of DCT

A four channel data acquisition system is developed to study somatosensory evoked potentials in the normal subjects and scoliotic patients. The somatosensory evoke potentials are of micro volts range and are extracted from noise, by averaging the responses obtained after multiple electrical stimulations. The electrical stimulations and data acquisition are carried out in synchronization and their parameters are controlled by a computer. The system have advantages like cost effectiveness, high sensitivity two signals and ability to record low amplitude complex patterns over existing data acquisition systems.

Southern Biomedical Engineering Conference - Proceedings

Development of an efficient data acquisition system for recording spinal somatosensory evoked potentials

A common feature in both VT and SVT a the occurrence of ventricular complexes with wide and large deflections, thereby masking the occurrence of the low amplitude P-waves. To identify the presente of atrio-ventricular (AV) dissociation, it becomes necessary to detect the P-waves. The work presented in this article focusses on the detection of the dissociated rhythm of P-waves based on the singular value ratio (SVR) spectrum of a residual signal constructed after subtracting the principal component from the original ECG record. The reliability of our method has been tested on 20 examples of wide QRS type arrhythmia waveforms and verified by comparing the reconstructed waveforms with the corresponding set of intra-atrial recordings.

Department of Electrical Engineering

Computers in Cardiology

Computer detection of P-waves from wide QRS type arrhythmia recordings

The enhancement of the P and T component waves especially in cases where the amplitude of the recorded noise exceeds that of the component waves is presented. A linear model is proposed for representation of the electrocardiogram (ECG) from a signal processing perspective whereby, enhancement of P and T waves is achieved by elimination of weak interfering components whose frequency range exceeds the fundamental repetition frequency of the recorded ECG. The algorithm requires a priori estimation of the QRS waveforms. The remainder electrogram (RECG) obtained is modeled as the output of an all-pole filter excited by the sequence resulting from the convolution of the wavelets and the impulse train.

Elimination of weak interfering components from electrocardiographic signals using linear prediction analysis

For the first time unstable periodic orbits(UPO's)are extracted from experimental human ECG signals for the quantitative classification of normal and abnormal cases. The normal cardiac system is characterized by three or four UPO's whereas pathological conditions such as premature ventricular contraction (PVC), Ventricular Fibrillation(VF), and Atrioventricular(AV) block shows different trends in the number and distribution of the UPO's.

Annual International Conference of the IEEE Engineering in Medicine and Biology - Proceedings

Dynamics of cardiac system through unstable periodic orbits

An algorithm is presented for identification of ventricular ectopic beats (PVC) by performing a beat by beat analysis of Electrocardiogram (ECG). The discrete Cosine Transform (DCT) of a windowed ECG cycle is decomposed into spectra of system and excitatory functions representing action potential and excitation pattern of the heart muscle during cardiac cycle. The Autoregressive (AR) modeling of the system function provides necessary information for identification of PVC's. The partial energy spectrum derived from the DCT coefficients characterizes the decay rate of DCT of the system function and is related to bandwidths of resonances in the AR spectrum. The Algorithm was able to successfully identify PVC's from the recordings of MIT-BIH database. Under noisy conditions, the Algorithm clearly distinguishes PVC patterns from those of normal beats upto a signal to noise ratio (SNR) of 10 dB.

Automatic detection of PVC's using autoregressive models

In this work, traditional, clinical methods of evaluating autonomic neural control systems are compared with the newer method based on the signal processing applied on Heart Rate Variability signal. The information available from the time-frequency representation of HRV signals is useful in classifying the subject population into four classes namely normal, sympathetic loss, parasympathetic loss and combined losses. The results suggest that the presence of autonomic neuropathy can be clearly detected by short time HRV signal analysis possible in the clinical environment.

Correlation of autonomic function tests to the HRV signal analysed using short time Fourier transform

Background: Prompted by the results of CAST results, attention has shifted from class I agents that primarily block sodium channels to class III agents that primarily block potassium channels for pharmacological management of certain cardiac arrhythmias. Recent studies demonstrated that sodium channel blockade, while antiarrhythmic at the cellular level, was inherently proarrhythmic in the setting of a propagating wave front as a result of prolongation of the vulnerable period during which premature stimulation can initiate reentrant activation. From a theoretical perspective, sodium (depolarizing) and potassium (repolarizing) currents are complementary so that if antiarrhythmic and proarrhythmic properties are coupled to modulation of sodium currents, then antiarrhythmic and proarrhythmic properties might similarly be coupled to modulation of potassium currents. The purpose of the present study was to exploit the role of repolarization currents during reentrant excitation. Methods and Results: To assess the generic role of repolarizing currents during reentry, we studied the responses of a two- dimensional array of identical excitable cells based on the FitzHugh-Nagumo model, consisting of a single excitation (sodium-like) current and a single recovery (potassium-like) current. Spiral wave reentry was initiated by use of S1S2 stimulation with the delay timed to occur within the vulnerable period (VP). While holding the sodium conductance constant, the potassium conductance (g(K)) was reduced from 1.13 to 0.70 (arbitrary units), producing a prolongation of the action potential duration (APD). When g(K) was 1.13, the tip of the spiral wave rotated around a small, stationary, unexcited region and the computed ECG was monomorphic. As g(K) was reduced, the APD was prolonged and the unexcited region became mobile (nonstationary), such that the tip of the spiral wave inscribed an outline similar to a multipetaled flower; concomitantly, the computed ECG became progressively more polymorphic. The degree of polymorphism was related to the APD and the configuration of the nonstationary spiral core. Conclusions: Torsadelike (polymorphic) ECGs can be derived from spiral wave reentry in a medium of identical cells. Under normal conditions, the spiral core around which a reentrant wave front rotates is stationary. As the balance of repolarizing currents becomes less outward (eg, secondary to potassium channel blockade), the APD is prolonged. When the wavelength (APD · velocity) exceeds the perimeter of the stationary unexcited core, the core will become unstable, causing spiral core drift. Large repolarizing currents shorten the APD and result in a monomorphic reentrant process (stationary core), whereas smaller currents prolong the APD and amplify spiral core instability, resulting in a polymorphic process. We conclude that, similar to sodium channel blockade, the proarrhythmic potential of potassium channel blockade in the setting of propagation may be directly linked to its cellular antiarrhythmic potential, ie, arrhythmia suppression resulting from a prolonged APD may, on initiation of a reentrant wave front, destabilize the core of a rotating spiral, resulting in complex motion (precession) of the spiral tip around a nonstationary region of unexcited cells. In tissue with inhomogeneities, core instability alters the activation sequence from one reentry cycle to the next and can lead to spiral wave fractionation as the wave front collides with inhomogeneous regions. Depending on the nature of the inhomogeneities, wave front fragments may annihilate one another, producing a nonsustained arrhythmia, or may spawn new spirals (multiple wavelets), producing fibrillation and sudden cardiac death.

Circulation

Proarrhythmic response to potassium channel blockade: Numerical studies of polymorphic tachyarrhythmias

Suppression of responses to premature stimulation has been the guiding principle in managing many cardiac arrhythmias. Recent clinical trails revealed that sodium channel blockade increased the incidence of re-entrant cardiac arrhythmias resulting in sudden cardiac death, although the physiologic mechanism remains uncertain. Potassium channel blockade offers an alternative mechanism for suppressing responses to premature stimuli. We have developed a simple model of a 2D sheet of excitable cells. We can initiate re-entrant activation with stimuli timed to occur within a period of vulnerability (VP). Reducing the Na conductance increases the VP while reducing the K conductance increases the collective instability of the array, and arrhythmias similar to torsades de pointes seen in patients subjected to K channel blocked can be readily initiated. Thus, while K channel blockade may suppress excitability by prolonging the action potential duration, it appears to simultaneously exhibit proarrhythmic properties that result in complex re-entrant arrhythmias.

Indian Journal of Physiology and Pharmacology

Potassium channel blockade amplifies cardiac instability numerical studies of torsades de pointes

Artificial neural networks have recently been introduced in the field of computer-based analysis of the electrocardiogram (ECG). The purpose of the present study was to examine the performance of a neural network in an ECG classification task. ECGs recorded from 272 patients with anterior myocardial infarction and 479 subjects without myocardial infarction were studied. Fifteen QRS measurements of the leads V2-V4 were used as inputs to the network. The network was trained using 502 ECGs. Thereafter, a comparison of network, conventional criteria and human expert was performed using a test set of249 ECGs. The neural network showed a higher sensitivity than the conventional criteria (79% versus 68%, p<0.01), both having a specificity of 97%. The performance of the human expert was the same as that of the neural network. In conclusion, it seems possible that neural networks could be used to improve the performance of some parts of ECG interpretation programs. © 1992 IEEE.

Proceedings - Computers in Cardiology, CIC 1992

Neural network versus electrocardiographer and conventional computer criteria in diagnosing anterior infarct from the ECG

The electrophysiology of the heart is represented by a Linear Time Invariant (LTI) system function to represent the electrocardiogram (ECG) in an efficient way. The EC G as well as its component waves, P, QRS and T, are independently subjected to Discrete Cosine Transform (DCT) and transform outputs are modelled using sixth order system for complete ECG and second order system for each component wave. When the complete ECG model is decomposed into three second order subsystems connected in parallel, denominator polynomials of these subsystems are closely matching with the denominator polynomials of three second order models derived for component waves. Sum of the component models impulse response is in good agreement with that of the ECG model. A relation between system parameters and the time domain signal peaks is presented. Validity of the method is discussed for the ECGs of bizarre shape and more than one ECG cycle in a sample frame is considered for modelling. The model provides a method to analyse and synthesise ECGs of different morphologies. © 1992 IEEE.

Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS

Digital model for electrocardiogram

Results on the modeling and synthesis of the electrocardiogram (ECG) from its component waves after a suitable transformation are presented. The transformed components add up to the transform of the complete beat and thus allow synthesis of the ECG. Pole-zero modeling of the transformed signals by the Shanks method proved very successful.