Natarajan Sriraam

Spectral entropy feature subset selection using SEPCOR to detect alcoholic impact on gamma sub band visual event related potentials of multichannel electroencephalograms (EEG)

Display Omitted The proposed feature selection method is based on maximization of class separability and minimum correlation between selected features.ICA is applied on Physionet alcoholic EEG database to separate pure EEG epochs from artifacts such as eye blink, body and cranial muscle movements.As per our knowledge, this is the first attempt to remove eye blink and muscle activity (> 30Hz) from the EEG epochs of the alcoholic EEG database.Results show better classification accuracy compared to previous studieson the same Physionet alcoholic EEG database.The normalized spectral entropy feature is used for extensive analysis in this study for the first time. The problem of analyzing and identifying regions of high discrimination between alcoholics and controls in a multichannel electroencephalogram (EEG) signal is modeled as a feature subset selection technique that can improve the recognition rate between both groups. Several studies have reported efficient detection of alcoholics by feature extraction and selection in gamma band visual event related potentials (ERP) of a multichannel EEG signal. However, in these studies the correlation between features and their class information is not considered for feature selection. This may lead to redundancy in the feature set and result in over fitting. Therefore in this study, a statistical feature selection technique based on Separability & Correlation analysis (SEPCOR) is proposed to select an optimal feature subset automatically that possesses minimum correlation between selected channels and maximum class separation. The optimal feature selection consists of a ranking method that assigns ranks to channels based on a variability measure (V-measure). From the ranked feature set of highly discriminative features, different subsets are automatically selected by heuristically applying a correlation threshold in steps from 0.02 to 0.1. These subsets are applied as input features to multilayer perceptron (MLP) neural network and k-nearest neighbor (k-NN) classifiers to discriminate alcoholic and control visual ERP. Prior to feature selection, spectral entropy features are computed in gamma sub band (30-55Hz) interval of a 61-channel multi-trial EEG signal with multiple object recognition tasks. Independent Component Analysis (ICA) is performed on raw EEG data to remove eye blink, motion and muscle artifacts. Results indicate that both classifiers exhibit excellent classification accuracy of 99.6%, for a feature subset of 22 optimal channels with correlation threshold of 0.1. In terms of computation time, k-NN classifier outperforms multilayer perceptron-back propagation (MLP-BP) network with 7.93s whereas MLP network takes 55s to perform the recognition task with the same accuracy. Compared to feature section methods used in previous studies on the same EEG alcoholic database, there is a significant improvement in classification accuracy based on the proposed SEPCOR method.

Features ranking for the classification of epileptic seizure from temporal EEG

Electroencephalogram shortly termed as EEG is considered as a fundamental segment for the assessment of epilepsy. EEG recordings, in turn, provide important information about epileptic se izure discharges and play an important role for the neurologist in defining the electroclinical disorders. The classification of epileptic seizures from normal EEG merely depends on the selection of appropriate EEG feature. The specific study emphasizes on the selection of best EEG features for classification and detection of epileptic seizure from temporal EEG. This study makes use of twelve different features to select suitable quantitative features. During the experimental study, it was observed that the variance, maximum, wavelet log energy entropy, rms and band power features were increasing parameters for epileptic EEG, on the other hand, the minimum, wavelet Shannon entropy and zero crossing features were increasing parameters for normal EEG and shows distinguishable differences between normal and epileptic EEG. Feature ranking test was applied using ttest, receiving operator characteristics, Bhattacharyya and Wilcoxon methods to select significant features. The eight features were considered favorable features for classification purpose whereas other four were not.

Comparative Study of Fuzzy Entropy with Relative Spike Amplitude Features for Recognizing Wake-Sleep Stage 1 EEGs

Electroencephalogram (EEG) based sleep stage analysis considered to be the gold standard method for assessment of sleep architecture. Of importance, transition between the first two stages, wake-sleep stage 1 found to be reliable quantitative tool for drowsiness and fatigue detection. The selection of appropriate feature pattern for EEGs is a quite challenging task due to its non-linear and non-stationary nature of the EEG signals. This research work attempts to provide a comparative study of time influence of time domain feature, relative spike amplitude (RSA) with entropy feature, fuzzy entropy(FE) for recognizing the transition between wake and sleep stage 1. EEGs extracted from offline polysomnography database is used and the extracted RSA and FE wake-sleep stage 1 derived EEG features are further classified using a feedback recurrent Elman neural network (REN) classifier. EEGs are segmented into 1s pattern. Simulation of the REN classifier revealed that the FE feature with REN yields a CA of 99.6% compared to that of with RSA feature. Comparative Study of Fuzzy Entropy with Relative Spike Amplitude Features for Recognizing WakeSleep Stage 1 EEGs

Characterization of EEG signals for identification of alcoholics using ANOVA ranked approximate entropy and classifiers

This work focuses on non-linear characterization of 61-channel electroencephalogram (EEG) signal for detecting alcoholics using ranked Approximate Entropy (ApEn) parameters. Significant channels that contribute to the detection of alcoholism are selected by ranking the ApEn features based on ANOVA test. In order to classify alcoholics from control, the ranked feature set is applied to two non-linear classifiers, namely Back Propagation Neural Network (BPNN) and Support Vector Machine (SVM) Classifiers respectively. The performance of the classifiers is evaluated in terms of classification accuracy as well as computational processing time. Experimental results reveal that the BPNN classifier with 40 hidden neurons and SVM classifier with a polynomial kernel of order 3 perform with an accuracy of 90% with only 32 ranked ApEn coefficients.

Analyzing thermal runaway in semiconductor devices using the constrained method of optimization

Thermal runaway is a major cause of failure of semiconductor devices in electronic systems. Analyzing the conditions for thermal runaway and its prevention is important to prevent this failure mechanism. In this paper, the Lagranges constrained method of optimization is applied to the problem of thermal runaway. Cases of thermal runaway in MOSFET,BJT and semiconductor ICs are discussed. A case of power generation and dissipation represented by a quadratic function in two-variables is taken as an example and the method of application is explained. The geometrical interpretation of the mathematical results is also discussed. Methods of prevention of thermal runaway in a few types of semiconductor components are given.

Classification of Breast Thermograms Using Statistical Moments and Entropy Features with Probabilistic Neural Networks

Breast cancer is considered as one of the life-threatening disease among woman population in developingaswellasdevelopedcountries.Thisspecificstudyreportsonclassificationofbreast thermogramsusingprobabilisticneuralnetwork(PNN)withfourstatisticalmomentsfeaturesmean, standarddeviation,skewnessandkurtosisandtwoentropyfeatures,ShannonentropyandWavelet packetentropy.TheCLAHEhistogramequalizationalgorithmwithuniformandRayleighdistributions wereconsideredforcontrastenhancementofbreastthermalimages.Theasymmetrydetectionwas performedbyapplyingbilateralratio.Atotalof95testimages(normal=53,abnormal=42)was considered.SimulationstudyshowsthatCLAHE-RDwithwaveletentropyfeaturesconfirmsthe existenceofsymmetryontherightandleftbreastthermalimages.Anoverallclassificationaccuracy of92.5%wasachievedusingtheproposedmultifeatureswithPNNclassifier.Theproposedtechnique thusconfirmsthesuitabilityasascreeningtoolforasymmetrydetectionaswellasclassificationof breastthermograms. KEywoRdS Breast Cancer, Classification, Entropy, Statistical Moments, Thermography

A pilot study on patch sensor based Photo Plethysmography (PPG) for heart rate measurements

The advent of Integrated Healthcare technology has helped in the development of portable medical devices for measuring vital physiological parameters such as such as Blood glucose and Blood pressure etc. These devices can be afforded by the common man and are readily available in the market. Medical devices of this kind give the firsthand information about the health condition of the patient and also suggests patient if any necessary actions need to be taken based on the observed symptoms. Although several attempts have been made to develop wearable non-invasive medical devices, still there is a huge demand to catch the need of resources constrained population. This paper suggests the design of a wearable patch sensor based on Photo Plethysmography (PPG) unit which can be used as a patch sensor or a wearable device for measuring the heart rate of the patient. The device consists of a sensor developed by Maxim Integrated Circuit for acquiring the PPG signal. The heart rate obtained from the PPG device is communicated serially to Arduino Processor using I2C serial communication, also accelerometer is embedded with the system for detecting undesirable hand movement while measuring heart rate from the signal derived from the PPG device. The Arduino Processor is also connected with the blue tooth model for further communication. The sensor can be placed either in cloth or in thin rubber sheet such that total unit can be tied around the wrist or can be designed place in any other part of the body were PPG signal can be recorded. From the preliminary pilot study, it can be concluded that the proposed patch sensor based PPG yields promising results for the measurement of heart rate. The device needs to be validated with large sample size before introducing to the clinical trial.

Recognition of infant cries using wavelet derived mel frequency feature with SVM classification

Infant cry is considered as primary source of information by the pediatric community to understand the discomfortness of full term babies. This fundamental acoustic signal establishes close connectivity between the mother and the baby. Attempts have been made to understand the infant cry pattern to exploit the hidden information useful for clinical investigation. This research study highlight the recognition of research study highlight the cry pattern from full term babies by making use of wavelet transform followed by Mel frequency estimation. Single level decomposition with symmetric discrete wavelet transform was applied on the preprocessed raw cry signal. For the experimental study, database from local hospital with a sample size of 60 was considers after due ethical clearance and three conditions namely hunger, pain and discomfort was considered. A Support Vector Machine (SVM) classifier with a linear kernel was employed to perform the binary classification. The simulation results reveal a classification accuracy of 93.09%, 90.27% and 71.29% for hunger-discomfort, pain-hunger and discomfort-pain respectively. It can be concluded that the proposed feature pattern classifier found to yield promising results suitable for clinical interpretation.

Study and investigation of continuous cardiac monitoring using vernier EKG with myRIO processor

In todays advancement in technology, affordable healthcare services still not in pace for constrained population. Especially cardiac-related screening needs special attention which increases the mortality rate of chronic cardiac patients every day. Several studies showed that most of the rural healthcare centre does not have the Electro Cardiogram (ECG) System for recording the physiological conditions of the patients. This specific study attempts to investigate the usage of Vernier EKG sensor with real time processor for developing portable ECG system. The preliminary work investigates the suitability of the ECG signal quantitatively and qualitatively for clinical support. A case study was shown with two feature extraction, namely, R-R interval, beats per minute and spectral entropy and the application of KNN classifier to distinguish two different physiological conditions. It was observed that the proposed configuration found to be a cost effective procedure. The quantitative and qualitative results confirm the suitability of the proposed system for real time cardiac monitoring of patients as well as screening tool.

Detection of alcoholic impact on visual event related potentials using beta band spectral entropy, repeated measures ANOVA and k-NN classifier

In this paper, the effect of alcohol on visual event related potentials (visual ERPs) of 61-channel electroencephalogram (EEG) is investigated using Spectral Entropy (SE) parameters in beta band (13–28Hz). The Repeated Measures Anova (RMANOVA) test is used to determine within group and between group variability of SE values in (61 channels) various specified regions of the brain during a visual single object recognition task. In order to discriminate visual ERPs produced in alcoholics and controls within the beta band, SE features are applied to a k-NN classifier. RMANOVA tests show that there is little statistical significance (p>0.05) as far as the marginal mean between groups is considered. However, the marginal mean variability within group indicates that there are some selected channels in which the statistical significance is observed (p<0.05). The classification results show that k-NN classification achieves 90.41% accuracy with k=8. It is also observed that the mean of spectral entropy values increase mainly in the left frontal region of the brain in alcoholics as compared to controls.