Ramakrishnan Swaminathan

Imperceptibility-Robustness tradeoff studies for ECG steganography using Continuous Ant Colony Optimization

ECG steganography is performed using DWT-SVD and quantization watermarking scheme.Imperceptibility-robustness tradeoff is investigated.Continuous Ant Colony Optimization provides optimized Multiple Scaling Factors.MSFs are superior to SSF in providing better imperceptibility-robustness tradeoff. ECG Steganography ensures protection of patient data when ECG signals embedded with patient data are transmitted over the internet. Steganography algorithms strive to recover the embedded patient data entirely and to minimize the deterioration in the cover signal caused by the embedding. This paper presents a Continuous Ant Colony Optimization (CACO) based ECG Steganography scheme using Discrete Wavelet Transform and Singular Value Decomposition. Quantization techniques allow embedding the patient data into the ECG signal. The scaling factor in the quantization techniques governs the tradeoff between imperceptibility and robustness. The novelty of the proposed approach is to use CACO in ECG Steganography, to identify Multiple Scaling Factors (MSFs) that will provide a better tradeoff compared to uniform Single Scaling Factor (SSF). The optimal MSFs significantly improve the performance of ECG steganography which is measured by metrics such as Peak Signal to Noise Ratio, Percentage Residual Difference, Kullback-Leibler distance and Bit Error Rate. Performance of the proposed approach is demonstrated on the MIT-BIH database and the results validate that the tradeoff curve obtained through MSFs is better than the tradeoff curve obtained for any SSF. The results also advocate appropriate SSFs for target imperceptibility or robustness.

Segmentation of Breast Tissues in Infrared Images Using Modified Phase Based Level Sets

In this study, segmentation of frontal breast tissues in infrared thermography is proposed using modified phase based level set method. The images considered for this work are obtained from open source database PROENG. An improved diffusion rate model is adopted and incorporated in distance regularized level set framework. Local phase information is used as an edge indicator for the evolution of level set function. Region based statistics and overlap measures are computed to compare and validate the segmented region of interests against ground truths. . Further, the obtained values are compared with the reported numerical values of three segmentation methods. The results show that the proposed level set method is able to extract the breast tissues in infrared images and able to address the inherent limitations in thermograms such as low contrast and absence of clear edges. A high amount of correlation between the segmented output and ground truths is observed. The performance of the proposed segmentation method is better when compared to reported segmentation methods. The adopted method seems to be effective in identifying the lower breast boundary and inflammatory folds present in breast thermograms.

Analyzing origin of multifractality of surface electromyography signals in dynamic contractions

The aim of this study is analyze the origin of multifractality of surface electromyography (sEMG) signals during dynamic contraction in nonfatigue and fatigue conditions. sEMG signals are recorded from triceps brachii muscles of twenty two normal healthy subjects. The signals are divided into six equal segments on time scale for normalization. The first and sixth segments are considered as nonfatigue and fatigue condition respectively. The source of multifractality can be due to correlation and probability distribution. The original sEMG series are transformed into shuffled and surrogate series. These three series namely, original, shuffled and surrogate series in nonfatigue and fatigue conditions are subjected to multifractal detrended fluctuation analysis (MFDFA) and features are extracted. The results indicate that sEMG signals exhibit multifractal behavior. Further investigation revealed that origin of multifractality is primarily due to correlation. The origin of multifractality due to correlation is quantified as 80% in nonfatigue and 86% in fatigue conditions. This method of multifractal analysis may be useful for analyzing progressive changes in muscle contraction in varied neuromuscular studies.

Classification of muscle fatigue using surface electromyography signals and multifractals

Muscle fatigue is commonly experienced in both normal and subjects with neuromuscular disorders. Surface electromyography (sEMG) signals are useful technique for analyzing muscle fatigue. sEMG signals are highly nonstationary and exhibit complex nonlinear characteristic in dynamic contractions. In this work, an attempt is made to classify sEMG signals recorded from biceps brachii muscles in nonfatigue and fatigue using multifractal features. The signals are recorded from 26 healthy normal adult subjects while performing standard experimental protocol involving dynamic contraction. The preprocessed signals are divided into six segments. The first and last segments are considered as nonfatigue and fatigue conditions respectively. The signals are then subjected to multifractal detrended moving average algorithm and eight multifractal features are extracted from both conditions. Further, information gain (IG) based ranking is used for reducing the number of features. Three different classification algorithms are employed namely, k-Nearest Neighbor algorithm (kNN), Naive Bayes (NB) and logistic regression (LR) for classification. The results show that signals exhibit multifractal characteristics and the multifractal features such as, generalized Hurst exponent, degree of multifractality and scaling exponent slope are significantly different in fatigue condition. The Hurst exponent for small fluctuation and degree of multifractality are found to be very highly significant feature. The LR and kNN classifier performance gave an accuracy of 84% and 82% respectively. This method of using multifractal features appears to be useful in classifying sEMG signals in dynamic contraction. This study can also be extended to classify fatigue condition in various neuromuscular disorders.

Multifractal analysis of sEMG signals for fatigue assessment in dynamic contractions using Hurst exponents

Multifractal analysis are useful to characterize complex physiological time-series. In this work, surface EMG signals recorded from biceps brachii muscles of 30 subjects are analyzed in dynamic fatigue conditions using multifractal techniques. The signals are segmented into six zones for time normalization. The first and last zones are considered as nonfatigue and fatigue conditions. The preprocessed signals are subjected to multifractal analysis and Hurst exponent function is computed. Three features, namely maximum and minimum exponent and strength of multifractality are used for analyzing nonfatigue and fatigue regions. The results indicate strength of multifractality is very high in fatigue condition and highly significant (p>2.7E-6) as compared to nonfatigue condition. The multifractal Hurst features are found to be useful in analyzing sEMG signal characteristics and this work can be extended for studying neuromuscular conditions.

Analyzing the influence of curl speed in fatiguing biceps brachii muscles using sEMG signals and multifractal detrended moving average algorithm

In this work, an attempt has been made to analyze surface electromyography (sEMG) signals of fatiguing biceps brachii muscles at different curl speeds using multifractal detrended moving average (MFDMA) algorithm. For this purpose, signals are recorded from fifty eight healthy subjects while performing curl exercise at their comfortable speed until fatigue. The signals of first and last curls are considered as nonfatigue and fatigue conditions, respectively. Further, the number of curls performed by each subject and the endurance time is used for computing the normalized curl speed. The signals are grouped into fast, medium and slow using curl speeds. The curl segments are subjected to MFDMA to derive degree of multifractality (DOM), maximum singularity exponent (MXE) and exponent length multifractality index (EMX). The results show that multifractal features are able to differentiate sEMG signals in fatiguing conditions. The multifractality increased with faster curls as compared with slower curl speed by 12%. High statistical significance is observed using EMX and DOM values between curl speed and fatigue conditions. It appears that this method of analyzing sEMG signals with curl speed can be useful in understanding muscle dynamics in varied neuromuscular conditions and sports medicine.

Analysis of surface EMG signals during dynamic contraction using Lempel-Ziv complexity

In this work, an attempt has been made to analyze progression of muscle fatigue in surface electromyography (sEMG) signals by estimating the complexity. The sEMG signals are acquired from biceps brachii of 50 healthy volunteers during dynamic contraction. The pre-processed signals are segmented into non-overlapping epochs of various sizes (500ms, 750ms and 1000ms) and Lempel-Ziv Complexity (LZC) is computed for each epoch. The linear regression technique is used to track the slope variations of LZC. The values of LZC show a decreasing trend during the progression of muscle fatigue. The magnitude of negative trend remained nearly constant irrespective of epoch size. Further, inter-subject variability of LZC measure is found to be minimum. The results shows that this method is useful in analyzing progression of muscle fatigue during dynamic contractions.

A Binary Bat Approach for Identification of Fatigue Condition from sEMG Signals

In this work, an attempt has been made to investigate the effectiveness of binary bat algorithm as a feature selection method to classify sEMG signals under fatigue and nonfatigue conditions. The sEMG signals are recorded from the biceps brachii muscle of 50 healthy volunteers. The signals are preprocessed and then multiscale Renyi entropy based feature are extracted. The binary bat algorithm is used for feature selection and the effectiveness is compared with information gain based ranker. The performance of the feature selection algorithms are validated by performing classification using Naive Bayes, and least square support vector machines. The results show a decreasing trend in the multiscale Renyi entropy with increase in scale. Additionally, higher entropy values where observed in fatigue condition. The classification results showed that a maximum accuracy of 86.66 % is obtained with least square SVM and binary bat algorithm. It appears that, this technique is useful in identifying muscle fatigue in varied clinical conditions.

Deep Learning on 1-D Biosignals: a Taxonomy-based Survey

Summary Objectives:  Deep learning models such as convolutional neural networks (CNNs) have been applied successfully to medical imaging, but biomedical signal analysis has yet to fully benefit from this novel approach. Our survey aims at (i) reviewing deep learning techniques for biosignal analysis in computer- aided diagnosis; and (ii) deriving a taxonomy for organizing the growing number of applications in the field. Methods:  A comprehensive literature research was performed using PubMed, Scopus, and ACM. Deep learning models were classified with respect to the (i) origin, (ii) dimension, and (iii) type of the biosignal as input to the deep learning model; (iv) the goal of the application; (v) the size and (vi) type of ground truth data; (vii) the type and (viii) schedule of learning the network; and (ix) the topology of the model. Results:  Between January 2010 and December 2017, a total 71 papers were published on the topic. The majority (n = 36) of papers are on electrocariography (ECG) signals. Most applications (n = 25) aim at detection of patterns, while only a few (n = 6) at predection of events. Out of 36 ECG-based works, many (n = 17) relate to multi-lead ECG. Other biosignals that have been identified in the survey are electromyography, phonocardiography, photoplethysmography, electrooculography, continuous glucose monitoring, acoustic respiratory signal, blood pressure, and electrodermal activity signal, while ballistocardiography or seismocardiography have yet to be analyzed using deep learning techniques. In supervised and unsupervised applications, CNNs and restricted Boltzmann machines are the most and least frequently used, (n = 34) and (n = 15), respectively. Conclusion:  Our key-code classification of relevant papers was used to cluster the approaches that have been published to date and demonstrated a large variability of research with respect to data, application, and network topology. Future research is expected to focus on the standardization of deep learning architectures and on the optimization of the network parameters to increase performance and robustness. Furthermore, application-driven approaches and updated training data from mobile recordings are needed.

Analysis of one repetition during biceps curl exercise among age-matched adult volunteers using endurance, curl speed and surface electromyography signals

Exercises under isometric and dynamic contractions are influenced by the rate coding and recruitment strategies. The study of muscle strength under dynamic contraction is normally performed using one-repetition maximum (1-RM) method. There are several variants of deriving one repetition method using number of repetitions and load that are useful in physical fitness and clinical rehabilitation program. However, the factors of dynamic contractions such as endurance time, speed of muscle contractions and muscle activity are not considered in 1-RM methods. The muscular activities are analyzed using surface electromyography (sEMG) signals. Limited work has been reported on the relationship between the 1-RM method and factors such as endurance time, speed of contraction and sEMG activity. In this work, a modified 1-RM method is proposed, namely, N-RM, using load, number of repetitions, endurance time, speed of contraction and normalized sEMG activity. For this purpose, sEMG signals are recorded from 58 healthy subjects under standard dynamic contraction protocol involving curl exercise. Conventional 1-RM is computed by using Epleys method and compared with proposed method using correlation analysis. The results show that 1-RM increases linearly with number of curls (r=1) but has a poor correlation coefficient with sEMG (r=0.01) and endurance time (r =0.4). The curl speed for lower 1-RM and higher 1-RM did not show any statistical difference (p =0.2). The proposed N-RM is observed to have good correlation with endurance time (r=0.734), curl speed (r=0.893) and sEMG activity (r=0.8851). These results demonstrate that the proposed N-RM is highly correlated to factors influencing the dynamic contractions. This method can be further extended to assess muscles under various clinical disorders and sports training.