Kayapanda M. Mandana

Application of principal component analysis to ECG signals for automated diagnosis of cardiac health

Electrocardiogram (ECG) is the P, QRS, T wave indicating the electrical activity of the heart. The subtle changes in amplitude and duration of ECG cannot be deciphered precisely by the naked eye, hence imposing the need for a computer assisted diagnosis tool. In this paper we have automatically classified five types of ECG beats of MIT-BIH arrhythmia database. The five types of beats are Normal (N), Right Bundle Branch Block (RBBB), Left Bundle Branch Block (LBBB), Atrial Premature Contraction (APC) and Ventricular Premature Contraction (VPC). In this work, we have compared the performances of three approaches. The first approach uses principal components of segmented ECG beats, the second approach uses principal components of error signals of linear prediction model, whereas the third approach uses principal components of Discrete Wavelet Transform (DWT) coefficients as features. These features from three approaches were independently classified using feed forward neural network (NN) and Least Square-Support Vector Machine (LS-SVM). We have obtained the highest accuracy using the first approach using principal components of segmented ECG beats with average sensitivity of 99.90%, specificity of 99.10%, PPV of 99.61% and classification accuracy of 98.11%. The system developed is clinically ready to deploy for mass screening programs.

APPLICATION OF HIGHER ORDER CUMULANT FEATURES FOR CARDIAC HEALTH DIAGNOSIS USING ECG SIGNALS

Electrocardiogram (ECG) is the electrical activity of the heart indicated by P, Q-R-S and T wave. The minute changes in the amplitude and duration of ECG depicts a particular type of cardiac abnormality. It is very difficult to decipher the hidden information present in this nonlinear and nonstationary signal. An automatic diagnostic system that characterizes cardiac activities in ECG signals would provide more insight into these phenomena thereby revealing important clinical information. Various methods have been proposed to detect cardiac abnormalities in ECG recordings. Application of higher order spectra (HOS) features is a seemingly promising approach because it can capture the nonlinear and dynamic nature of the ECG signals. In this paper, we have automatically classified five types of beats using HOS features (higher order cumulants) using two different approaches. The five types of ECG beats are normal (N), right bundle branch block (RBBB), left bundle branch block (LBBB), atrial premature contraction (APC) and ventricular premature contraction (VPC). In the first approach, cumulant features of segmented ECG signal were used for classification; whereas in the second approach cumulants of discrete wavelet transform (DWT) coefficients were used as features for classifiers. In both approaches, the cumulant features were subjected to data reduction using principal component analysis (PCA) and classified using three layer feed-forward neural network (NN) and least square-support vector machine (LS-SVM) classifiers. In this study, we obtained the highest average accuracy of 94.52%, sensitivity of 98.61% and specificity of 98.41% using first approach with NN classifier. The developed system is ready clinically to run on large datasets.

Data mining approach for coronary artery disease screening

Coronary artery disease (CAD) is the major cause of mortality in the world. Although there is a significant level of advancement in medical science and technology, this disease still remains challenging to the common people. The aim of this study is to develop a computer assisted screening system that will help early detection of CAD and improved patient management with the limited resources in the developing countries. The present system is developed from an initial marked data set. Ten risk factors have been investigated for the risk stratification of CAD. Two decision tree models -ID3 and CART, have been applied for finding a preliminary set of rules from the annotated database. The extracted rules have been clinically validated by a group of cardiologists as per their medical experience and acumen in finding a final set of rule base. The dataset used for automatic generation of model consists of 500 subjects. The present screening system provides risk stratification for CAD based on easily available medical data and it produces rules that can be easily interpreted by the medical experts. The developed system is ready to clinically validate on a large dataset.

Identifying coronary artery disease from photoplethysmogram

This paper presents the idea of a non invasive screening system for identifying Coronary Artery Disease (CAD) patients from fingertip Photoplethysmogram (PPG) signal. A combined feature set, related to heart rate variability (HRV) as well as shapes of PPG waveform has been defined for distinguishing CAD and non CAD subjects. Support Vector Machine (SVM) is used for classification. Our methodology yields sensitivity and specificity scores of 0.82 and 0.88 respectively in identifying CAD patients on a corpus of 112 subjects, selected from MIMIC II dataset. Further, we achieved sensitivity and specificity scores of of 0.73 and 0.87 on another dataset of 30 subjects, collected from an urban hospital using commercial oximeter device.

An unsupervised learning for robust cardiac feature derivation from PPG signals

We propose here derivation algorithms for physiological parameters like beat start point, systolic peak, pulse duration, peak-to-peak distance related to heart rate, dicrotic minima, diastolic peak from Photoplethysmogram (PPG) signals robustly. Our methods are based on unsupervised learning mainly following morphology as well as discrete nature of the signal. Statistical learning has been used as a special aid to infer most probable feature values mainly to cope up with presence of noise, which is assumed to be insignificant compared to signal values at each investigation window. Performance of the proposed method is found to be better than other standard methods, yielding precision and sensitivity more than 97% obtained from three real life data sets.We propose here derivation algorithms for physiological parameters like beat start point, systolic peak, pulse duration, peak-to-peak distance related to heart rate, dicrotic minima, diastolic peak from Photoplethysmogram (PPG) signals robustly. Our methods are based on unsupervised learning mainly following morphology as well as discrete nature of the signal. Statistical learning has been used as a special aid to infer most probable feature values mainly to cope up with presence of noise, which is assumed to be insignificant compared to signal values at each investigation window. Performance of the proposed method is found to be better than other standard methods, yielding precision and sensitivity more than 97% obtained from three real life data sets.

CAD Patient Classification Using MIMIC-II

With availability of large volume of collected data from healthcare centers and significant improvement in computation power, evidence based learning is helping in building robust disease diagnostic models.

Automated in vivo delineation of lumen wall using intravascular ultrasound imaging

This paper presents a novel methodology for automated detection and extraction of the lumen wall from Intravascular Ultrasound (IVUS) frames. IVUS is an in-vivo pull back imaging technique and provides a sequential frame of images for diagnosis of atherosclerotic heart disease. The detection and segmentation of lumen wall is necessary for predicting the arterial wall blockage. Lumen wall is recognized and segmented with the help of seed refinement and random walks algorithms, in tunica and lumen area. The proposed methodology was tested on 147 frames of 13 patients. Proposed method achieves significant performances for automated lumen wall detection and extraction as compared with existing literature.

iCarMa: Inexpensive Cardiac Arrhythmia Management -- An IoT Healthcare Analytics Solution

Ubiquity of smartphones with array of inbuilt sensors, pave ways to inexpensive mobile-health systems, particularly for cardio-vascular health monitoring. Smartphones, wearable sensors, and body area sensors play an important role as a part of Internet of Things (IoT) m-health ecosystem. In this paper, we present iCarMa to enable an inexpensive auto-triggered arrhythmia cardiac management solution catering the need of in-house, round-the-clock cardiac health monitoring. It facilitates early detection of fatal cardiac conditions like asystole, extreme bradycardia, extreme tachycardia, ventricular flutter and ventricular tachycardia, which often compel an individual to get admitted in Intensive Care Unit (ICU). Smartphone or wearable sensor extracted photoplethysmogram (PPG) is the sole physiological signal that is considered to characterize the cardiac anomalous events. Our main novelty is to precisely detect and remove the motion artifacts in PPG signals and to ensure accuracy in arrhythmia condition detection, specifically to reduce the false negative alarms. We establish the efficacy of proposed solution, iCarMa by large set of experiments with field-collected and MIT-Physionet PPG signals.

Noise detection in smartphone phonocardiogram

This paper presents a demo proposal of a standalone smartphone application that can automatically analyse the signal quality of PCG, as it is recorded on a low-cost smartphonebased digital stethoscope. Features, related to the inherent pattern of the autocorrelated signal envelope, have been used for classifying and discarding the noisy portions from a continuous PCG. Our application has been successfully deployed on Nexus 5 and tested on several clean and noisy PCG signals with sensitivity 78.91% and specificity 70.83%

Novel Statistical Post Processing to Improve Blood Pressure Estimation from Smartphone Photoplethysmogram

Blood pressure (BP) is considered to be an important biomarker for cardiac risk estimation. This paper deals with a non-conventional way of estimating BP using smartphone captured Photoplethysmogram (PPG) that enables unobtrusive health monitoring at home for possible alert generation. We have proposed a set of features that are independent to the inbuilt sensor of the capturing device. It is also observed that, BP estimated from a typical smartphone PPG signal fluctuates in successive cardiac cycles due to poor signal quality compared to a medical grade device. Hence, a novel post processing block is introduced, that rejects data depending on the BP distribution over all cardiac cycles in a session. Finally, Half Range Mode is used as a statistical average for the accepted sessions. This post processing methodology outperforms standard statistical averages in providing a better representative BP per session. The methodology yields mean absolute errors of 7.4% and 9.1% for predicting systolic and diastolic pressure respectively when validated over a dataset with a wide variation of BP.