Farbod Hosseyndoust Foomany

Real-time electrogram analysis for monitoring coronary blood flow during human ventricular fibrillation: Implications for CPR

BACKGROUND Effective chest compressions during prolonged ventricular fibrillation (VF) have been shown to increase the chances of successful defibrillation to a rhythm associated with a sustainable cardiac output. There is currently no effective method of recording the degree of antegrade coronary artery flow during chest compression in VF. OBJECTIVE This study sought to quantify the relationship between the antegrade coronary flow and the characteristics of human VF using near real-time wavelet-based electrocardiographic markers. METHODS VF experiments were conducted in 8 isolated human hearts. The Langendorff perfusion enabled different flow rates (perfusion) during VF, which allowed for the simulation of chest compression with different efficacies. After the initiation of VF, the hearts were maintained in ischemia (no flow) for 3 minutes, followed by a 2-minute reperfusion and defibrillation. The experiments were repeated at flows of 0%, 30%, and 100% of baseline perfusion, and volume-conducted surface electrograms were recorded and analyzed using continuous wavelet transform in 5-second frames. RESULTS Near real-time wavelet features were derived that demonstrated significant differences in the multicomponent nature of VF signals and predicted perfusion rate characteristics for different flow rates (i.e., 0%, 30%, and 100%; P < .0006). A pattern classifier was trained using the feature values from 5 hearts, and the flow rates for 3 additional hearts were predicted with an accuracy of 90%. CONCLUSION VF electrogram characteristics as measured by wavelet analysis relate to antegrade coronary flow rate during VF. These findings suggest that chest compression efficacy of physiological importance could be monitored using near real-time wavelet analysis.

Wavelet-based markers of ventricular fibrillation in optimizing human cardiac resuscitation

During cardiac resuscitation from ventricular fibrillation (VF) it would be helpful if we could monitor and predict the optimal state of the heart to be shocked into a perfusing rhythm. Real-time feedback of this state to the emergency medical staff (EMS) could improve the survival rate after resuscitation. In this paper, using real world out-of-the-hospital human VF data obtained during resuscitation by EMS personnel, we present the results of applying wavelet markers in predicting the shock outcomes. We also performed comparative analysis of 5 existing techniques (spectral and correlation based approaches) against the proposed wavelet markers. A database of 29 human VF tracings was extracted from the defibrillator recordings collected by the EMS personnel and was used to validate the waveform markers. The results obtained by the comparison of the wavelet based features with other spectral, and correlation-based features indicates that the proposed wavelet features perform well with an overall accuracy of 79.3% in predicting the shock outcomes and hence demonstrate potential to provide near real-time feedback to EMS personnel in optimizing resuscitation outcomes.

Fusion of structural and functional cardiac magnetic resonance imaging data for studying Ventricular Fibrillation

Magnetic Resonance Imaging (MRI) techniques such as Current Density Imaging (CDI) and Diffusion Tensor Imaging (DTI) provide a complementing set of imaging data that can describe both the functional and structural states of biological tissues. This paper presents a Joint Independent Component Analysis (jICA) based fusion approach which can be utilized to fuse CDI and DTI data to quantify the differences between two cardiac states: Ventricular Fibrillation (VF) and Asystolic/Normal (AS/NM). Such an approach could lead to a better insight on the mechanism of VF. Fusing CDI and DTI data from 8 data sets from 6 beating porcine hearts, in effect, detects the differences between two cardiac states, qualitatively and quantitatively. This initial study demonstrates the applicability of MRI-based imaging techniques and jICA-based fusion approach in studying cardiac arrhythmias.

Classification of music instruments using wavelet-based time-scale features

Separation of sounds from different sources plays a significant role in success of auditory scene analysis and multimedia content recognition. In this paper, we propose wavelet-based features for discrimination of signals from various music instruments. One hundred and fifty-two music segments from thirteen different instruments were selected from a public music database (Universitat Pompeu Fabra). We performed automatic instrument classification of segments from 13 instruments using selected wavelet features which resulted in accuracy as high as 85%. The wavelet features, along with the considerations suggested and elaborated on here, while are successful for solving the problem at hand, could be applied to many signal processing problems in other domains.

Feature-based MRI data fusion for cardiac arrhythmia studies

Current practices in studying cardiac arrhythmias primarily use electrical or optical surface recordings of a heart, spatially limited transmural recordings, and mathematical models. However, given that such arrhythmias occur on a 3D myocardial tissue, information obtained from such practices lack in dimension, completeness, and are sometimes prone to oversimplification. The combination of complementary Magnetic-Resonance Imaging (MRI)-based techniques such as Current Density Imaging (CDI) and Diffusion Tensor Imaging (DTI) could provide more depth to current practices in assessing the cardiac arrhythmia dynamics in entire cross sections of myocardium. In this work, we present an approach utilizing feature-based data fusion methods to demonstrate that complimentary information obtained from electrical current distribution and structural properties within a heart could be quantified and enhanced. Twelve (12) pairs of CDI and DTI image data sets were gathered from porcine hearts perfused through a Langendorff setup. Images were fused together using feature-based data fusion techniques such as Joint Independent Component Analysis (jICA), Canonical Correlation Analysis (CCA), and their combination (CCA+jICA). The results suggest that the complimentary information of cardiac states from CDI and DTI are enhanced and are better classified with the use of data fusion methods. For each data set, an increase in mean correlations of fused images were observed with 38% increase from CCA+jICA compared to the original images while mean mutual information of the fused images from jICA and CCA+jICA increased by approximately three-fold. We conclude that MRI-based techniques present potential viable tools in furthering studies for cardiac arrhythmias especially Ventricular Fibrillation.

Analysis of reliability metrics and quality enhancement measures in current density imaging

Low frequency current density imaging (LFCDI) is a magnetic resonance imaging (MRI) technique which enables calculation of current pathways within the medium of study. The induced current produces a magnetic flux which presents itself in phase images obtained through MRI scanning. A class of LFCDI challenges arises from the subject rotation requirement, which calls for reliability analysis metrics and specific image registration techniques. In this study these challenges are formulated and in light of proposed discussions, the reliability analysis of calculation of current pathways in a designed phantom and a pig heart is presented. The current passed is measured with less than 5% error for phantom, using CDI method. It is shown that Gausss law for magnetism can be treated as reliability metric in matching the images in two orientations. For the phantom and pig heart the usefulness of image registration for mitigation of rotation errors is demonstrated. The reliability metric provides a good representation of the degree of correspondence between images in two orientations for phantom and pig heart. In our CDI experiments this metric produced values of 95% and 26%, for phantom, and 88% and 75% for pig heart, for mismatch rotations of 0 and 20 degrees respectively.

Noise distribution and denoising of current density images

Abstract. Current density imaging (CDI) is a magnetic resonance (MR) imaging technique that could be used to study current pathways inside the tissue. The current distribution is measured indirectly as phase changes. The inherent noise in the MR imaging technique degrades the accuracy of phase measurements leading to imprecise current variations. The outcome can be affected significantly, especially at a low signal-to-noise ratio (SNR). We have shown the residual noise distribution of the phase to be Gaussian-like and the noise in CDI images approximated as a Gaussian. This finding matches experimental results. We further investigated this finding by performing comparative analysis with denoising techniques, using two CDI datasets with two different currents (20 and 45 mA). We found that the block-matching and three-dimensional (BM3D) technique outperforms other techniques when applied on current density (J). The minimum gain in noise power by BM3D applied to J compared with the next best technique in the analysis was found to be around 2 dB per pixel. We characterize the noise profile in CDI images and provide insights on the performance of different denoising techniques when applied at two different stages of current density reconstruction.

A novel approach to quantification of real and artifactual components of current density imaging for phantom and live heart.

Spatial distribution of injected current in a subject could be calculated and visualized through current density imaging (CDI). Calculated CDI paths however have a limited degree of accuracy due to both avoidable methodological errors and inevitable limitations dictated by MR imaging constraints. The source and impact of these limitations are scrutinized in this paper. Quantification of such limitations is an essential step prior to passing any judgment about the results especially in biomedical applications. An innovative technique along with metrics for evaluation of range of errors using baseline and phase cycle MR images is proposed in this work. The presented approach is helpful in pinpointing the local artifacts (areas for which CDI results are suspect), evaluation of global noises and artifacts and assessment of the effect of approximation algorithms on real and artifactual components. We will demonstrate how this error/reliability evaluation is applicable to interpretation of CDI results and in this framework, report the CDI results for an artificial phantom and a live pig heart in Langendorff setup. It is contended here that using this method, the inevitable trade-off between details and approximations of CDI components could be monitored which provides a great opportunity for robust interpretation of results. The proposed approach could be extended, adapted and used for statistical analysis of similar methods which aim at mapping current and impedance based on magnetic flux images obtained through MRI.

Blind source separation in characterizing ECG pre-shock waveforms during ventricular fibrillation

Ventricular Fibrillation (VF) is a cardiac arrhythmia for which the only available treatment option is defibrillation by electrical shock. Existing literature indicates that VF could be the manifestation of different sources controlling the heart with different degrees of organization. In this work we test the hypothesis that the pre-shock waveforms of successful and unsuccessful shock outcomes could be related to the number of independent sources present in these waveforms. The proposed method uses Blind Source Separation (BSS) to extract independent components in frequency direction from a pig database consisting of 20 pre-shock waveforms. The slope of the energy capture curve was used as an indicator to demonstrate the number of independent sources required to model the pre-shock waveforms. The results were also quantified by performing a linear discriminant analysis based classification achieving an overall classification accuracy of 75%. The results indicate that successful cases can be modeled with less number of independent sources compared to unsuccessful cases.

Modeling Current Density Maps Using Aliev–Panfilov Electrophysiological Heart Model

Most existing studies of cardiac arrhythmia rely on surface measurements through optical or electrical mapping techniques. Current density imaging (CDI) is a method which enables us to study current pathways inside the tissue. However, this method entails implementation complexities for beating ex vivo hearts. Hence, this work presents an approach to simulate and study the current distributions in different cardiac electrophysiological states. The results are corroborated by experimental data, and they indicate that different states were distinguishable. The CDI simulations can be used for studying cardiac arrhythmias under simulation conditions which are otherwise impossible or difficult to be implemented experimentally.