Master of Science in Engineering, University of Akron, 2015, Biomedical Engineering

An important driving force in the human circulatory system is the heart. The electrical activity of the heart can be recorded as P, Q, R, S and T waves that constitute an electrocardiogram (ECG) signal. The main method used in this thesis involves the discrete Hermite Transform (DHmT) that provides shape analysis for ECG signals. Earlier work proved that the DHmT method of characterizing ischemia from ECGs was fast and accurate for the ECG signals from precordial lead V4. Myocardial ischemia is a lack of oxygen flow to heart tissue exhibited in ECGs through different shapes. This thesis extends that analysis to limb lead MLIII to characterize ischemia. When used with other leads, analysis of MLIII confirms ischemic characteristics for particular arteries. This work is important because cardiologists tend to characterize such events using different standard ECG leads. The ECG signal shapes can be characterized by a new digital signal transform that is shape-based: the discrete Hermite transform (DHmT), method discovered by Mugler et al. (2000). An online archived database (http://physionet.org) is used for obtaining ECG waveforms from the European ST-T databank. These signals are computationally analyzed using MATLAB programming. The DHmT and related methods are applied to quickly and accurately identify characteristics of ischemia in ST segments. Heartbeats that show artifacts or abnormalities are not included in the analysis. The speed of the computation is such that a 2-hour ECG recording can be processed in 2.19 minutes on a standard PC. It tailors the analysis to the individual patient. The analysis concentrates on MLIII to determine the ischemic characteristic shapes of ST depression and elevation. The results show that high sensitivity, specificity, and positive predictive value are given by this method when applied to ischemic episodes. It is believed that this research into MLIII is the first of its kind.

Committee: Dale Mugler Dr. (Advisor); Narender Reddy Dr. (Committee Member); Yang Liu Dr. (Committee Member) Subjects: Biomedical Engineering; Biomedical Research

Doctor of Philosophy (PhD), Wright State University, 2010, Computer Science and Engineering PhD

Wearable Health Monitoring Systems (WHMS) have drawn a lot of attention from the research community and the industry during the last decade. The development of such systems has been motivated mainly by increasing healthcare costs and by the fact that the world population is ageing. In addition to that, RandD in WHMS has been propelled by recent technological advances in miniature bio-sensing devices, smart textiles, microelectronics and wireless communications techniques. These portable health systems can comprise various types of small physiological sensors, which enable continuous monitoring of a variety of human vital signs and other physiological parameters such as heart rate, respiration rate, body temperature, blood pressure, perspiration, oxygen saturation, electrocardiogram (ECG), body posture and activity etc. As a result, and also due to their embedded transmission modules and processing capabilities, wearable health monitoring systems can constitute low-cost and unobtrusive solutions for ubiquitous health, mental and activity status monitoring. The majority of the currently developed WHMS research prototypes and products provide the basic functionality of continuously logging and transmitting physiological data. However, WHMS have the potential of achieving early detection and diagnosis of critical health changes that could enable prevention of health hazardous episodes. To do that, they should be able to learn individual user baselines and also employ advanced information processing algorithms and diagnostics in order to discover problems autonomously and detect alarming health trends, and consequently, inform medical professionals for further assistance. In an effort to advance the capabilities of a wearable system towards these goals, we focus in this dissertation on the development of a novel WHMS, called Prognosis. The developed prototype platform includes the following innovative features, which constitute the main research contributions of this wor (open full item for complete abstract)

Committee: Nikolaos Bourbakis PhD (Advisor); Soon Chung PhD (Committee Member); Yong Pei PhD (Committee Member); Arnab Shaw PhD (Committee Member); Larry Lawhorne PhD (Committee Member) Subjects: Computer Science; Engineering; Health Care; Information Systems

Master of Computer and Information Science, Cleveland State University, 2025, Washkewicz College of Engineering

This paper presents E2P, a biometric key generation protocol designed for secure pairing IoT wearable devices using the RR intervals from EEG and ECG signals. While EEG and ECG differ significantly in their sensing mechanisms—capturing brain and cardiac activity, respectively—physiological correlations between both signals, particularly through RR intervals, can be exploited to establish a shared secret. The proposed protocol addresses several technical challenges, including signal noise, temporal misalignment, and differences in amplitude and waveform structure. A zero-phase FIR filter is applied to remove noise while preserving the timing of key features. The Teager–Kaiser Energy Operator (TKEO) is employed to enhance the detection of relevant cardiac components in EEG signals prior to peak identification. RR intervals are then extracted from both signals using peak detection with outlier removal based on statistical thresholds. These intervals are quantized using the LLoyd-Max algorithm and encoded with Gray code to reduce bit transition sensitivity. To reconcile mismatches between the two sources, BCH error correction is applied, enabling consistent key generation. Experimental evaluations confirm the feasibility of cross-modal key generation and demonstrate resistance to impersonation and replay attacks. E2P introduces a new approach to multi-modal biometric security by leveraging the intrinsic relationship between brain and heart activity. It offers a lightweight solution applicable to resource-constrained wearable systems, and establishes a foundation for future work involving additional biosignals and adaptive processing techniques.

Committee: Ye Zhu (Committee Chair); Zicheng Chi (Committee Member); Tianyun Zhang (Committee Member) Subjects: Computer Science

Master of Science in Engineering, University of Akron, 2023, Mechanical Engineering

Human factors concerns with automation have emerged as contributing factors in many aviation accidents in the past few decades. Adaptive automation, where a system dynamically assigns tasks to automation or the pilot based on workload, has been proposed as a potential solution to many of these concerns. This study examines how one proposed method of adaptive automation, using physiological data to measure workload, could be implemented using machine learning. Electrocardiogram (ECG), electrodermal activity (EDA), and facial electromyography (fEMG) data was collected at both low and high workload while subjects completed common tasks performed by pilots. This data was used to train binary classification neural networks, with many models achieving high accuracy. The models were then applied to different data with varying workload, achieving poor results. The results of this study identify design requirements for adaptive automation systems using this method, and further study required for practical application.

Committee: Chen Ling (Advisor); Shengyong Wang (Committee Member) Subjects: Aerospace Engineering; Engineering

Doctor of Philosophy, University of Akron, 2023, Electrical Engineering

This work describes a fully analog ECG motion artifacts (MAs) and baseline wander elimination circuit that can be incorporated in the analog sensor frontend. In the proposed method, the R-peaks, as useful components of ECG signals, are detected by a high pass filter (HPF) and excluded from the moving average input. By linearly interpolating the down-sampled moving average output, the baseline wander can be effectively detected. The final output is generated by subtracting the extracted baseline wander from the corrupted ECG waveforms. Owing to various types of switch capacitor integrated circuits including the Biquad filter, integrator and double sampling sample and hold (S/H) circuits for realizing DC offset and abrupt changes removal, HPF and moving average, linear interpolation and delay chain circuits, this method can be implemented fully on-chip. Therefore, the power consumption and chip area are drastically reduced compared to existing schemes, thus, can be suitable for using in long-term ECG monitoring devices. The proposed algorithm is implemented on the single chip (utilizing 0.18-μm CMOS technology with 1.8-V power supply) and verified by on-body testing over 24 subjects within the age group of 10 to 55 for different types of motion artifacts due to various activities including walking, texting, sleeping, and intentionally touching the skin electrodes. The measurement results show signal-to-noise ratio (SNR) improvement of almost 16-dB and on average 10% improvement in delta percentage root-mean-square difference (ΔPRD), where the power consumption of the chip is only 6.6-μW with core area of 0.85×2.16 mm2.

Committee: Kye-Shin Lee (Advisor); Huu Nghi Tran (Committee Member); Ronald Otterstetter (Committee Member); Jae-won Choi (Committee Member); Igor Tsukerman (Committee Member) Subjects: Electrical Engineering

Doctor of Philosophy, Case Western Reserve University, 2019, EECS - Computer Engineering

With the onset of the big data era, designing efficient and secure machine learning frameworks to analyze large-scale data is in dire need. This dissertation considers two machine learning paradigms, the centralized learning scenario, where we study the secure outsourcing problem in cloud computing, and the distributed learning scenario, where we explore blockchain techniques to remove the untrusted central server to solve the security problems. In the centralized machine learning paradigm, inference using deep neural networks (DNNs) may be outsourced to the cloud due to its high computational cost, which, however, raises security concerns. Particularly, the data involved in DNNs can be highly sensitive, such as in medical, financial, commercial applications, and hence should be kept private. Besides, DNN models owned by research institutions or commercial companies are their valuable intellectual properties and can contain proprietary information, which should be protected as well. Moreover, an untrusted cloud service provider may return inaccurate and even erroneous computing results. To address above issues, we propose a secure outsourcing framework for deep neural network inference called SecureNets, which can preserve both a user's data privacy and his/her neural network model privacy, and also verify the computation results returned by the cloud. Specifically, we employ a secure matrix transformation scheme in SecureNets to avoid privacy leakage of the data and the model. Meanwhile, we propose a verification method that can efficiently verify the correctness of cloud computing results. Our simulation results on four- and five-layer deep neural networks demonstrate that SecureNets can reduce the processing runtime by up to 64%. Compared with CryptoNets, one of the previous schemes, SecureNets can increase the throughput by 104.45% while reducing the data transmission size by 69.78% per instance. We further improve the privacy level in SecureNets and implement (open full item for complete abstract)

Committee: Pan Li (Advisor); Loparo Kenneth (Committee Member); An Wang (Committee Member); Ayday Erman (Committee Member) Subjects: Computer Engineering

PHD, Kent State University, 2018, College of Arts and Sciences / Department of Computer Science

Heart diseases are the major causes of morbidity and fatality in senior age group which affect their productivity and lifestyle significantly. ECG is a noninvasive means of maintaining healthy heart. One of the major abnormalities of heart is arrhythmia that consists of irregular heartbeats due to ectopic nodes. Currently available systems lack sufficient accuracy and finer real time classification, which affects the treatment. In this research, machine learning and parallelization techniques have been developed for the real-time analysis of ECG for diagnosing the finer classes of arrhythmias and irregular heartbeats. An integrated approach combining Markov model and bivariate Gaussian distribution has been proposed for an integrated analysis of the temporal and morphological features. Area subtraction techniques have been proposed for detecting the embedded waveforms. The analysis has been extended with a look-ahead pattern analysis algorithm for identifying different classes of irregular beats. The execution efficiency has been further improved to accommodate diagnosis of other heart-diseases in real-time by exploiting GPU based SIMT parallelism that performs beat level analysis concurrently. The implementation results show very high accuracy.

Committee: Arvind Bansal Dr. (Advisor); Javed Khan Dr. (Committee Member); Cheng Chang Lu Dr. (Committee Member); Gokarna Sharma Dr. (Committee Member); Jeffery Nielson Dr. (Committee Member) Subjects: Artificial Intelligence; Computer Science

Doctor of Philosophy, The Ohio State University, 2014, Comparative and Veterinary Medicine

Goal: Primary goal of this study was to determine if exercise training or exposure to drugs would protect rats from acute and reversible heart failure produced by IV imipramine. Methods: Fifty-four, male, Sprague-Dawley rats were allocated randomly into 5 intervention groups (n = 10): (1) sedentary, (2) exercise, (3) carvedilol, (4) clenbuterol, and (5) dobutamine. Six rats in each intervention received imipramine challenge. A 6th group (n = 4) was sedentary but received a matched-volume vehicle challenge. The variables studied were: orthogonal lead ECGs, systemic arterial and left ventricular pressures, maximal rates of rise and fall of left ventricular pressure, left ventricular end-diastolic and end-systolic volumes, body weight, and weights of brain, heart, and adrenal. Values were expressed as means ± SE. ECGs were obtained (1) after rats had received interventions and while anesthetized with pentobarbital (baseline pre-surgery), then ECGs and hemodynamics were recorded (2) before they received imipramine or vehicle infusion (baseline instrumentation), (3) at the mid-point of infusion (mid-dose), (4) at the end-point of infusion (end-dose), and (5) 1 hour after cessation of infusion (end recovery). Differences of statistical significance in means for all parameters measured during imipramine challenge were assessed among groups and times using 2-way ANOVA with repeated measures. Results and Discussion: Exercise, carvedilol, clenbuterol, and dobutamine produced physiological effects consistent with their known properties. All rats survived imipramine challenges with hemodynamic and ECG changes typical of acute imipramine exposure, i.e., an initial decrease in function with or without spontaneous recovery during infusion, and then recovery nearly complete within 1 hour after cessation of infusion. No intervention altered statistically (i.e., blunted or exaggerated) the hemodynamic responses to imipramine; however differences among interventions were noted. (open full item for complete abstract)

Committee: Robert Hamlin DVM, PhD (Advisor); Carl Leier MD (Committee Member); Mark Ziolo PhD (Committee Member); Steven Devor PhD (Committee Member) Subjects: Pharmacology; Physiology

Doctor of Philosophy, Case Western Reserve University, 2014, EECS - Electrical Engineering

The recording and analysis of electroencephalography (ECG) plays crucial roles in clinical research and diagnosis. As a result, the development of automatic ECG analysis algorithms has been rapidly growing in recent decades. However, conventional ECG analysis encounters tradeoff between computational cost and performance accuracy. This study aims to develop a series of real-time (online) ECG analysis algorithms that include heartbeat detection and ECG arrhythmia classification. We first propose a novel phase space based method for heartbeat detection that maps the ECG data into a two-dimensional reconstructed attractor. Unlike conventional algorithms, our detector replaces the preprocessing stage with a reconstruction process. This improvement highly reduces the computational cost. Moreover, we introduce a two-dimensional decision mechanism in order to obtain high performance accuracy at the detection stage. For the ECG arrhythmia classification study, an unsupervised classification algorithm referred to as “superparamagnetic clustering” is introduced to ECG analysis field for the first time. Current studies in ECG classification mainly use supervised artificial intelligence methods. The common drawbacks of these classifiers include: they are incapable of discriminating clusters with significant population differences; manual annotation efforts by clinicians/researchers are required in order to form the training sets; the network training procedure is computationally expensive. The proposed arrhythmia ECG classifier overcomes these issues because of the non-parametric configuration of the classifier. Clustering with different desirable discrimination levels could be realized by adjusting the “temperature” parameter. Moreover, this study explicitly involves exploration of the feature selection issue. To ensure the most reliable configuration, an appropriate number of the most significant features should be selected from the candidate pool. A comparative study betwee (open full item for complete abstract)

Committee: Kenneth Loparo (Committee Chair); Thomas Dick (Committee Member); Frank Jacono (Committee Member); Samden Lhatoo (Committee Member); Farhad Kaffashi (Committee Member) Subjects: Biomedical Engineering; Health Care; Information Technology; Mathematics; Statistics

Doctor of Philosophy, University of Akron, 2013, Biomedical Engineering

Coronary Artery Disease (CAD) and Obstructive Sleep Apnea (OSA) are both complex and significant clinical problems. The pathophysiological mechanisms that associate these diseases together are complicated. OSA can influence the broad spectrum of conditions caused by CAD, from subclinical atherosclerosis to myocardial infarction. Research in the last decade has shown a growing evidence of OSA as a risk factor for CAD [55]. The Sleep Heart Health Study (SHHS) was motivated by the increasing recognition of the frequent occurrence of sleep-disordered breathing in the general population and mounting evidence that sleep-disordered breathing may increase risk for a lot of cardiovascular diseases and may reduce quality of life [51]. Identification of early symptoms of ischemic events in the general population with or without OSA will act as a useful tool in providing an insight into when an intervention would be needed. Early identification of cardiac problems in this population may also lead to better treatment options at a reduced cost. In this current study we developed an algorithm for ischemic feature identification and its relation to occurrence of sleep disordered breathing. This algorithm uses the fast-shape based Hermite analysis to identify features in an electrocardiogram. It identifies the three different manifestations of ischemia in an ECG on a beat by beat basis: namely ST elevation, ST depression and T wave inversion. It further provides analysis of these features based on episodes of ischemic events during different stages of sleep and its association with physiological parameters like saturated percentage of oxygen (SpO2) and Heart rate. Quantitative analysis of the different ST segment features mentioned above were performed in subjects with and without OSA and during different stages of sleep. Significant differences were observed in features of ST segment depression corresponding to rise in severity of OSA in subjects in the control group (tho (open full item for complete abstract)

Committee: Dale Mugler Dr. (Advisor); Narender Redd Dr. (Committee Member); George Giakos Dr. (Committee Member); Joan Carletta Dr. (Committee Member); Kevin Kreider Dr. (Committee Member); Daniel Sheffer Dr. (Committee Member) Subjects: Biomedical Research

MS, University of Cincinnati, 2003, Engineering : Computer Engineering

In any type of signal processing, it has been demonstrated that it is important to remove noise from the signal before recognizing or classifying the patterns. Otherwise, the whole process may give wrong results. In this work the choice of denoising mechanisms for various types of input data and Gaussian noise is explored, to increase the signal strength. In this thesis, denoising the input signals using a wavelet transform is discussed. It is shown that the performance of a signal classifier improves when these denoising techniques are introduced before actually applying the classifier. For our experiments, the classifier applied is a hybrid intelligent system that employs three important techniques of artificial intelligence, namely genetic algorithms, neural networks and fuzzy logic. Along with explaining the denoising algorithm clearly, this work shows the importance of selection of a suitable wavelet for the given input data and thus shows that the efficiency of a signal denoiser depends on three factors: the thresholding techniques, the kind of wavelet used in denoising, and the synchronization between the wavelet selected and the input data. This statement is justified with results from experiments on ECG data which employ different kinds of wavelets such as Haar, Daubechies, Symlet and Coiflet. The improvements in denoising after using vector quantization of wavelet coefficients before thresholding are also discussed.

Committee: Dr. Carla Purdy (Advisor) Subjects:

Doctor of Philosophy, The Ohio State University, 2011, Veterinary Biosciences

The assessment of the electrocardiogram (ECG) liabilities is a key component in both clinical medicine and safety-pharmacology. However, despite the known circadian dependence of both pro-arrhythmic substrates and prevalence of arrhythmias, the optimal time for the ECG evaluation remains undetermined. In fact the time of day of the recording is seldom considered in the interpretation. This study assessed circadian changes in ECG parameters in a well-defined telemetered canine model. Diurnal differences were sought between normal dogs and dogs with failing hearts (i.e., with reduced ejection fraction and elevated NTproBNP) but not in heart failure (i.e., asymptomatic). Methods of analyzing ECGs from (the relatively few) dogs with clinical heart failure are not of great concern, since dogs so afflicted have clear echocardiographic and other clinical evidence of disease. So this study was directed more at (the much greater number of) dogs with subclinical disease, where standard methods are more equivocal, difficult to perform, and expensive. Healthy male dogs (n = 9) and male dogs with failing hearts (n=10) were instrumented for telemetered ECG recordings at rest (1-hour epochs) during periods of low (2AM) and high autonomic activation (6PM and 6AM); heart rates (HR), indices of cardiac conduction (PQ, QRS) and repolarization duration (QT, QTcF) as well as heart rate variability (RRSD), repolarization instability (QTSTV) and steepness of restitution (QT/TQ) were evaluated at each time-point. Data are mean ± SD and were compared (ANOVA). The number of dogs entered into this study was determined a priori to produce a power of ~0.8 to detect 15% differences in parameters having coefficients of variation of 25% with an alpha of 0.05. In both groups, at 2 AM, heart rates were slower and more variable while the PQ interval was longer, indicating strong parasympathetic control. No notable circadian differences in ventricular conduction (QRS duration) were found in either n (open full item for complete abstract)

Committee: Robert Hamlin (Advisor); Krista La-Perle (Committee Member); Carlos Couto (Committee Member); Mark Strauch (Committee Member) Subjects: Veterinary Services

Master of Education, Cleveland State University, 2012, College of Education and Human Services

The use of a dry electrode (DE), which does not rely on electrolytic solution, may circumvent potential disadvantages of a wet electrode (WE). The accuracy of the electrocardiogram (ECG) signal, provided by the electrode, is vital. The purpose of the study was to investigate if differences in signal quality, reflected by the signal-to-noise ratio (SNR), existed between the standard gel 3MTM Red DotTM 2560 electrode (WE) and the Orbital Research Incorporated (ORI) dry electrode (DE) over a 96 hour period of continuous wear. Assessments were made within electrode types, comparing potential signal deterioration within the electrode over time, and also between the two electrode types, comparing SNR over time. Twenty healthy adult volunteers completed the research protocol, each simultaneously wearing the two pairs of electrodes for 96 hours continuously in a lead II configuration. ECG tracings were collected simultaneously on different telemetry channels once a day over five consecutive days. The collection period consisted of six, three minute stages. The six stages included two bouts of rest, supine and standing, followed by three submaximal exercise stages, ending with one stage of standing rest. Data collected using the telemetry unit was de-noised by MatlabTM using sixth order Daubechies wavelet transform technology. No significant differences existed within the DE SNRs over time, indicating that SNR deterioration did not occur. Although a significant difference existed within the WE SNR between day 0 and 1 (17.83 ± 2.62 vs. 18.68 ± 2.35 on day 0 and day 1, respectively, p < .01) in the standing stage, the noise was reduced; therefore, SNR deterioration did not occur. The only significant difference (p < .01) between the WE and DE SNRs occurred on day 2 (19.94 ± 2.11 vs. 18.73 ± 1.97) and day 4 (20.16 ± 2.16 vs. 18.96 ± 2.21) in the supine stage, favoring the WE. The difference observed could be attributed to a potential loss of skin contact when in the supin (open full item for complete abstract)

Committee: Ken Sparks PhD (Committee Chair); Kathleen Little PhD (Committee Member); Michael Loovis PhD (Committee Member) Subjects:

Master of Science in Engineering, University of Akron, 2007, Biomedical Engineering

Increased attention to patient safety, demands for innovation in medical education, and accelerating advances in diagnostic and therapeutic procedures have all promoted a growing interest in the use of simulators for medical training and assessment.The current study proposed and developed a method for approximating and reproducing physiological signals in a programmable simulator using wavelet filtering. This method employed the technique of designing a template from an already existing source data,which then forms the basis of this realistic artificial biomedical signal generator/simulator. The simulated physiological signals included an electrocardiogram,blood pressure, respiratory signal, time derivative of thoracic impedance (dZ/dt), and photopletheysmogram. Templates were also designed for conditions exhibiting premature ventricular contraction, ventricular flutter and left bundle branch block in an electrocardiogram. The software was designed in MATLAB®, and DATAQ® Instruments DI-720 data acquisition system was used to display the simulated output. Evaluation of this simulator model was done both in quantitative and qualitative terms. The results proved that using wavelets for reconstruction of physiological signals minimized distortion and retained the significant features in each signal that was simulated.

Committee: Bruce Taylor (Advisor) Subjects: Engineering, Biomedical