Biomedical subjects

N Kumaravel

Publications and source records attributed to N Kumaravel.

11 recordsLinked to original sources

PubMed2003

Internet based expert system for the management of gallstones, renal, ureteric and bladder calculi.

An Internet based expert system for the management of gallstones, Renal, Ureteric and bladder calculi based on ultrasound images is presented in this paper. Calculi are due to abnormal collection of certain chemicals like oxalate, phosphate and Uric acid. These calculi can be present in kidney, Ureter or in Urinary bladder and also in gall bladder. The expert system is designed to assist the physician to detect, extract, classify and diagnose calculi with greater accuracy. It also helps physicians in the management of calculi based on the etiological analysis of calculi. The Expert system takes an ultrasound image as input along with the symptoms of the patients. The expert system extracts the renal calculi and analyzes it using different image processing techniques to extract the image features like size, location and texture. These image features along with the clinical data of the patient enable the expert system to provide the decisions to decide the future course of treatment with more accuracy.

Artificial Intelligence↗

PubMed2003

A wavelet coefficient smoothened RLS adaptive denoising model for ECG.

This paper mainly concentrate on Noise Cancellation methodology for bio-signals. In this paper we have smoothened the Wavelet based signal coefficient and adapted to the MSE using RLS Algorithm. We have applied the Noisy ECG to After reconstruction model (ARM) and Before reconstruction model (BRM) and is implemented and tested. The proposed model combines the advantage of Wavelet Transform (co-efficient smoothening) and Adaptive Filter. This new proposed model performed better and faster. This find many application in the filed of noise elimination.

Algorithms↗

PubMed2002

Segmentation of renal calculi in ultrasound images.

An algorithm proposed by Sridhar and Kumaravel is extended to include a framework for the detection of renal calculi. Calculi occur due to abnormal collection of certain chemicals like oxalate, phosphate and uric acid. These calculi can be present in the kidney, ureter or urinary bladder. Performance analysis is done to a set of five known algorithms using parameters such as success rate in calculi detection, border error metric and time. The framework is constructed by combining the best algorithm based on the performance analysis and a procedure to validate the detected calculi using the shadow it casts in ultrasound images. Ultrasound images of 37 patients are used for testing the algorithm. The detected calculi based on the framework match those determined by expert clinicians in more than 95% of the cases.

Algorithms↗

PubMed2001

Automatic segmentation of medical images for renal calculi and analysis.

Development of an automated system to identify renal calculi based on its physical characteristics is proposed. Calculi are due to abnormal collection of certain chemicals like oxalate, phosphate and uric arid. Renal calculi may be present in kidney, ureter or in urinary bladder. An algorithm is proposed to detect calculus using its shadow. The system also extracts the properties of calculi such as size, shape and location, which are vital for reliable diagnosis. This method has been implemented in the MATLAB/IDL platform and a considerable success rate is obtained.

Algorithms↗

PubMed2000

Optimal autoregressive model based medical image compression using genetic algorithm.

Image compression is the process of removal of redundant data in order to achieve savings in storage and communication costs. A new method for predictive coding of images is proposed using two-dimensional multiplicative autoregressive models. Genetic algorithm is implemented to compute autoregressive parameters. Comparison with other multiplicative autoregressive model based image compression method has also been reported here.

Algorithms↗

PubMed1999

Real time implementation of genetic algorithm cancellation of sinusoidal noise in ECG using TMS320C50 DSP processor.

Power-line interference (either 50 Hz or 60 Hz) is a significant source of noise in biomedical recording. Elimination of power-line interference in the Electrocardiogram (ECG) signal by various methods have been proposed in the past. Despite the high common mode rejection ratio of the input amplifiers of the Electrocardiographs, recordings are often contaminated by power-line (mains) frequency interference, usually phase shifted with respect to the mains voltage. The use of conventional filter affects the signal components adjacent to the mains line frequency. A variety of adaptive filters for this purpose have been proposed but they require some adaptation period whenever there is change in the power-line interference. This paper describes a method of implementing the genetic algorithm noise cancellation in real time, using TMS320C50 based add-on DSP card that does the process of conversion and subtraction and the host PC doing the training process.

Algorithms↗

PubMed1998

Genetic-algorithm cancellation of sinusoidal powerline interference in electrocardiograms.

The paper describes a method, based on a genetic algorithm, to remove sinusoidal powerline interference in electrocardiograms. There is a report on the use of the genetic algorithm to remove powerline interference for two different types of interference, powerline interference with frequency drift, and interference with frequency drift as well as third- harmonic distortion. The studies are conducted on electrocardiograms with simulated interference and also on actual noisy electrocardiogram records. The results obtained using the genetic algorithm in these cases of interference are presented.

Algorithms↗

PubMed1997

ECG compression and recognition using complete tree representation.

This paper deals with a novel method of Electrocardiogram (ECG) representation, compression and recognition using Complete Tree algorithm. The construction of Complete Tree is like overlaying a ECG waveform on a grid. The waveform is observed through the grid structure with vertical and horizontal grid lines. In addition to nodes created for each interval enclosed by that waveform on a quantisation level, nodes are also created for each point where the waveform cuts the vertical grid line between the current quantisation level and the next quantisation level. A leaf node represents the data sample of ECG at that position. For reconstruction of this ECG waveform, only the leaf nodes are used resulting in ECG compression. This representation is also useful in ECG pattern matching and recognition without reconstructing the original waveform. The ECGs with sampling rate of 500 sps are used. A compression ratio (CR) of 4.8:1 with percent RMS difference (PRD) of 8.45% is obtained. The reconstruction of original ECG waveform from its tree representation shows high fidelity in all its complexes of ECG. The results are compared with all other compression techniques such as AZTEC, TP, FAN, DPCM etc. By using Tree merging and splitting algorithms, the matching and recognition of ECG is implemented.

Algorithms↗

PubMed1997

Equivalent tree representation of electrocardiogram using genetic algorithm.

Electrocardiogram (ECG) gives the electrical activity of the heart. The number of data points required to represent the ECG signal is reduced by using a complete-tree representation. This reduced data structure (ECG Tree) is obtained by fitting the ECG signal in a grid structure consisting of both horizontal and vertical lines. The leaf nodes are the points where the vertical grid lines intersect with the ECG signal. These leaf nodes now form the features of the ECG signal. Some of these leaf nodes may be redundant and hence the reduction in the number of leaf nodes and thus optimization of the tree (equivalent tree) is done using a novel technique based on the Genetic Algorithm (GA). In this work, the equivalent tree is formed using GA consisting of four stages. First, from the group of generated leaf nodes various combinations of strings are constructed to form the population. Second, the fitness function is taken as the measure of the vertical distances between two neighbouring leaf nodes in order to evaluate the population with respect to their fitness values. Third, the selection procedure is used to give offsprings based on an assigned threshold value. Finally, crossover and mutation operations are performed repeatedly till an optimized population is obtained. The optimal nodes represent the equivalent tree. The Backpropagation Neural Network as a classifier is used to test the efficacy of the GA in this optimization problem.

Algorithms↗

PubMed1995

Integrating the ECG power-line interference removal methods with rule-based system.

The power-line frequency interference in electrocardiographic signals is eliminated to enhance the signal characteristics for diagnosis. The power-line frequency normally varies +/- 1.5 Hz from its standard value of 50 Hz. In the present work, the performances of the linear FIR filter, Wave digital filter (WDF) and adaptive filter for the power-line frequency variations from 48.5 to 51.5 Hz in steps of 0.5 Hz are studied. The advantage of the LMS adaptive filter in the removal of power-line frequency interference even if the frequency of interference varies by +/- 1.5 Hz from its normal value of 50 Hz over other fixed frequency filters is very well justified. A novel method of integrating rule-based system approach with linear FIR filter and also with Wave digital filter are proposed. The performances of Rule-based FIR filter and Rule-based Wave digital filter are compared with the LMS adaptive filter.

Algorithms↗

PubMed1994

Automatic diagnosis of neuro-muscular diseases using neural network.

An automatic diagnostic tool for neuromuscular diseases, based on the feature extraction and classification of myoelectric patterns using neural network is described. Electromyogram (EMG) signals are extracted from the patients during maximal contraction using needle electrodes. This EMG signal is digitized at a rate of 1000 samples/second. The myoelectric signal is divided into many time segments. Five time domain features are extracted from each of these segments and are averaged over the segments to obtain one feature set. This is applied to the neural network for classification. Results are presented for the diagnosis of polymyositis.

Electromyography↗