H. K. Verma

ANN-based QRS-complex analysis of ECG

Reliable detection the QRS complex in either a normal or an abnormal ECG and its analysis is the first and foremost task in almost every ECG signal analysis system aimed at the diagnostic interpretation of ECG. Conventionally, detection of the QRS complex is accomplished using a rule-based/algorithmic approach. This work, uses the learn and generalize approach of an artificial neural network (ANN) for the detection of QRS complexes in either a normal or an abnormal ECG. This is followed by the analysis of the QRS complex to designate and measure the morphological components within the QRS complex in all 12 standard leads. An ANN has been developed to detect the QRS complex in ECG and trained, with the help of back propagation algorithm, on more than a hundred ECGs selected from the CSE Data Set-3. The trained ANN was tested on all the recordings of the CSE Data Set-3 and the sensitivity has been found to be 99.11%. Subsequent to the identification of the QRS complex, an analysis of this complex and measurement of peak amplitudes of the component waves is done. The results are validated using the CSE multilead measurement results. Both the QRS detection and the QRS analysis software developed in C-language have been successfully implemented on a PC-AT. The results are found to be in agreement with visual measurements carried out by medical experts.

Reduction of blocking artifacts in JPEG compressed images

Image compression is a very important issue for several applications in the area of multimedia communications, the objective being reduction of storage and transmission costs. Many efficient coding techniques have been developed for various applications. Amongst them, the Joint Photographic Experts Group (JPEG) has been recommended for compression of continuous tone still images. However, the reconstructed images from JPEG compression produce annoying blocking artifacts near block boundaries, particularly in highly compressed images, because each block is transformed and quantized independently. Several techniques/algorithms have been proposed by researchers, both in spatial and frequency domains, for reduction of these artifacts with varied degree of success. These are briefly overviewed here. A new technique working in frequency domain, is proposed here by authors. This paper puts forth a method and an algorithm, working in frequency domain, for the detection and reduction of such blocking artifacts. These artifacts are modeled here as 2-D step functions between two neighboring blocks. Presence of the blocking artifacts is detected by using block activity based on human visual system (HVS) and block statistics. The boundary regions between blocks are identified as either smooth or non-smooth regions. The blocking artifacts in smooth regions are removed by modifying a few DCT coefficients appropriately, whilst an edge-preserving smoothing filter is applied to the non-smooth regions, i.e., genuine edges. The algorithm has been applied to variety of JPEG compressed images and results are compared with other postprocessing algorithms. The reduction in the blocking artifacts for each image have been evaluated using three indices, namely peak signal-to-noise ratio (PSNR), mean structure similarity (MSSIM) index based on human visual perception, and a new index, called here block boundary measure (BBM), applied to both vertical and horizontal block boundaries. The results show that the proposed method is very effective in detecting and reducing the blocking artifacts in JPEG compressed images.

Artificial neural network based wave complex detection in electrocardiograms

Abstract This paper describes a Predictive Neural Network (PNN) based technique to detect QRS complexes of electrocardiograms (ECGs). The PNN is trained, using the back propagation algorithm, on non-QRS portions of the ECG to predict the signal one-step ahead. High prediction error is then taken as an indication of the occurrence of a QRS complex. A simple peak detection logic is then invoked to mark the exact location and magnitude of either a Q- or an R- or an S- peak within the QRS complex. The performance of the detector software is illustrated with examples representative of different QRS morphologies. The accuracy of QRS complex detection has been tested using bipolar standard limb leads of a standard ECG library; a sensitivity of 98·96% has been achieved. A brief discussion on how well this technique performs in comparison with the other QRS detectors is also presented.

Direct data compression techniques for ECG signals: effect of sampling frequency on performance

The existing techniques for data compression can be divided into three main groups, namely direct, transformation and parameter extraction methods. The present paper deals with the direct data compression (DDC) methods as applied to ECG data. The performance has been evaluated on the basis of compression ratio, percent-root-mean-square difference and fidelity of the reconstructed signal. Further, in order to know the extent to which the diagnostic information is preserved during compression, peak and boundary measurements have been made both on the reconstructed and original ECG signal and compared. The objective of the present paper is to report the effect of sampling frequency on the aforementioned parameters as studied by the authors. The experimental results show that with the increase in sampling frequency the compression ratio increases and the percent-root-mean-square difference generally decreases. Further, the reconstructed signal is of higher quality having larger bandwidth and higher resolution...

ECG data compression using fast Walsh transform and its clinical acceptability

Abstract Data compression techniques can be grouped into three categories, direct and transformation compression and parameter extraction. The principles and details of ECG data compression implementation using fast Walsh transform are presented. Performance evaluation has been made on the basis of compression ratio and visual comparison. To know the extent to which the clinical information preserved in the reconstructed signal, peak and boundary measurements were made on both the reconstructed and original signal, and then compared. As the number of fast Walsh transform coefficients are reduced, the compression ratio increases. At higher compression ratios deviations in the R-peak are found to be higher than those in the other ECG peaks. Peak and boundary measurements reveal clinical acceptability of the compression algorithm, because the errors are quite tolerable. It is also worth noting that artefacts such as electromyographic noise are better eliminated because of filtering, and this also makes the s...

Signal-analysis and a heuro-logistic interpretation of multi-lead electrocardiograms

This paper describes a personal computer (PC) based analysis of multi-lead electrocardiograms (ECGs) and their interpretation employing an approach which is a combination of the two basic and conventional approaches, namely the heuristic approach and the logistic approach. The ECG analysis part of the software has been validated using the multi-lead ECGs of the common Standards in quantitative electrocardiography (CSE) database (Data Set 3). Using the spatial velocity approach, the analysis software reliably detects the QRS complexes and then the other component waves (i.e. P & T waves). More than 90% of the fiducial locations of various waves (i.e. P-on. P-off, QRS-on, QRS-off and T-end) estimated by the analysis software using this approach are found to be well within the tolerances recommended by the CSE. A successful attempt has been made to evaluate the parameters of diagnostic importance and interpret the multi-lead ECG analysis results using the heuro-logistic approach. The computerized interpretat...

Ultrasonic imaging system for flaw characterization

Abstract The ultrasonic imaging system presented here is useful for non-destructive evaluation of metal specimens. Based on the pulse-echo technique, it detects the presence, depth and shape of a flaw or discontinuity in specimens. The surface of the test specimen is scanned by an ultrasonic probe driven by a two-stepper motor assembly configured to get movements along x and y directions independently. An 8-bit microprocessor controls the stepper motors and measures the depth of the flaw at every position of the probe. The data on x and y coordinates of each probed point and depth of flaw at that point (as z coordinate) are stored in a read-write memory in real time. The recorded information is later displayed on a general-purpose CRO. In addition to the description of the system, its advantages, limitations and usage are also given in the paper.

ECG delineation techniques and applications in Cobra venom interacted patients

ECG signal provides a dynamic assessment of the cardiac activities. To analyze the ECG signal, some innovative ECG delineation techniques alongwith a few known techniques were used to provide the optimum strategy for marking of events. The techniques were applied on 125 records of CSE DS-3 database for calibration and subsequently, on the data acquired from no known-cardiac Cobra bitten patients under strict ethical protocol and medical supervision.