Hilmi Dajani - Academia.edu (original) (raw)

Papers by Hilmi Dajani

IEEE journal of biomedical and health informatics, Jan 27, 2016

Non-invasive blood pressure (BP) measurement is an important tool for managing hypertension and c... more Non-invasive blood pressure (BP) measurement is an important tool for managing hypertension and cardiovascular disease. However, automated non-invasive BP measurement devices, which are usually based on the oscillometric method, do not always provide accurate estimation of BP. It has been found that change in arterial stiffness (AS) is an underlying mechanism of disagreement between an oscillometric BP monitor and a sphygmomanometer. This problem is addressed by incorporating parameters related to AS in the algorithm for BP measurement. Pulse transit time (PTT) is first used to estimate AS parameters, which are fixed into a model of the oscillometric envelope. This model can then be used to perform curve fitting to the measured signal using only four parameters: systolic BP, diastolic BP, mean BP, and lumen area at zero transmural pressure. The proposed technique is independent of the experimentally-determined characteristic ratios that are commonly used in existing oscillometric me...

IEEE International Workshop on Medical Measurement and Applications, 2010

MeMeA 2010 is the fifth edition of a successful international workshop series in the field of mea... more MeMeA 2010 is the fifth edition of a successful international workshop series in the field of measurements in medicine and health care, with previous editions held in: Cetraro-Italy (2009), Ottawa-Canada (2008), Warsaw-Poland (2007), and Benevento-Italy (2006).

Canadian Acoustics, Sep 1, 2007

institute o

Canadian Acoustics, Sep 1, 2009

A desirable feature of modern hearing aids is the ability to automatically adjust its behavior in... more A desirable feature of modern hearing aids is the ability to automatically adjust its behavior in different acoustic environments. There are two kinds of approaches that can achieve this. One is based on "classification" of environments, another is "direct learning" of preferred hearing aid settings. Our focus is on the second approach, in which an artificial neural network (ANN) learns the preferred volume setting of the hearing aid user. The performance of such a direct-learning system strongly depends on the chosen signal features. While a large number of features have been derived for environment classification , we are now aware of any that have been derived specifically for "direct learning", which has different requirements. For example, environment classification should not in general be sensitive to the sound volume, whereas direct learning of volume setting not only depends on the volume, but also depends on how this volume affects speech intelligibility and user comfort. Moreover, the preferred volume setting depends on the hearing loss profile, and whether it is profound, severe, or moderate. The goal of this work is to derive suitable features that the ANN will use to set the volume such that it optimizes speech intelligibility. New features are proposed, which are based on measures of speech intelligibility, namely the Speech Intelligibility Index (SII) (ANSI S3.5-1997) and the Coherence SII (CSII) . The performance of these features is then investigated using a simulator of a hearing aid user (SPOT, 2009).

2008 IEEE International Workshop on Medical Measurements and Applications, 2008

ABSTRACT

4th International Workshop on Soft Computing Applications, 2010

Feed-Forward Neural Network (FFNN) has recently been utilized to estimate blood pressure (BP) fro... more Feed-Forward Neural Network (FFNN) has recently been utilized to estimate blood pressure (BP) from the oscillometric measurements. However, there has been no study till now that consolidated the role played by the different neural network (NN) training algorithms in affecting the BP estimates. This paper compares the estimation errors in the BP due to ten different training algorithms belonging to three classes: steepest descent (with variable learning rate, with variable learning rate and momentum, resilient backpropagation), quasi-Newton (Broyden-Fletcher-Goldfarb-Shanno, one step secant, Levenberg-Marquardt) and conjugate gradient (Fletcher-Reeves update, Polak-Ribiére update, Powell-Beale restart, scaled conjugate gradient) that are used to train two separate NNs: one to estimate the systolic pressure and the other one to estimate the diastolic pressure. The different training algorithms are compared in terms of estimation error (mean absolute error and standard deviation of error) and training performance (training time and number of training iterations to reach the optimal weights). The NN-based approach is also compared with the conventional maximum amplitude algorithm.

Pitch discrimination experiments have demonstrated that human listeners can detect very small fre... more Pitch discrimination experiments have demonstrated that human listeners can detect very small frequency changes in stimuli of short duration. Inspired by this ability, an algorithm for high resolution time-varying spectral analysis is proposed. Mathematical analysis, with various types of synthetic modulated signals, demonstrates that the proposed method correctly demodulates these signals. The resulting spectrogramlike display, referred to as a 'Fine Structure Spectrogram', shows the fine structure of the modulations in higher detail than is possible with conventional spectrograms. With recorded speech samples, the fine structure spectrogram detects small frequency and amplitude modulations in the formants of speech. It also appears to identify additional components in speech that are not detected by other methods.

The American journal of physiology

The purpose of this study was to characterize oscillations in muscle sympathetic nerve activity (... more The purpose of this study was to characterize oscillations in muscle sympathetic nerve activity (MSNA) in the frequency domain in healthy subjects and patients with congestive heart failure (CHF) and to relate these to blood pressure (BP), heart rate (HR), and breathing frequency. MSNA burst frequency was significantly greater in CHF [52 +/- 21 (n = 12) vs. 35 +/- 11 (n = 19) bursts/min, P < 0.05], whereas breathing frequency and HR were similar. There was no significant difference between CHF and healthy subjects in total power, harmonic power, and nonharmonic power in the MSNA spectrum from 0 to 0.5 Hz, but low frequency power (LF, 0.05-0.15 Hz, P < 0.05) was reduced in heart failure patients. There was less coherence between BP and MSNA in the LF range, but similar spectral power in both groups in the very LF (VLF, 0-0.05 Hz) and high frequency (0.15-0.5 Hz) ranges. The transfer of MSNA oscillations into BP in the VLF (P < 0.05) and LF (P < 0.02) ranges was significantly lower in CHF, but gains in the transfer function and in the coherence between BP and MSNA and in the coherence between respiration and MSNA were similar in the two groups. These observations indicate that modulation of MSNA by the arterial baroreflex and respiration is preserved in CHF. The loss of LF power in the MSNA signal may be due to impaired neuroeffector transduction. The higher sympathetic nerve firing rate in CHF would therefore appear to be caused by factors other than the loss of regulation by these two inhibitory influences.

The Journal of the Acoustical Society of America, 2015

2015 IEEE International Symposium on Medical Measurements and Applications (MeMeA) Proceedings, 2015

IEEE Transactions on Instrumentation and Measurement, 2015

An accurate noninvasive estimation of mean arterial pressure (MAP) is of great importance in the ... more An accurate noninvasive estimation of mean arterial pressure (MAP) is of great importance in the evaluation of circulatory function and prognosis of some cardiovascular diseases. This paper proposes a novel oscillometric MAP estimation method based on the dependence of pulse transit time (PTT) on cuff pressure (CP). The PTT computed as the time interval between the electrocardiogram (ECG) R-peaks and the maximum slope points on the oscillometric pulses is mathematically modeled by considering the cuff-arm-artery system and the blood flow dynamics. It is then analytically shown that MAP can be approximated as the CP at which the PTT is maximum. Based on our theoretical findings, a new method of MAP estimation from simultaneous ECG and oscillometric blood pressure measurements is proposed. Our proposed method is validated with a pilot study in which 150 recordings from 10 subjects are analyzed. The reference MAP is computed from the systolic and diastolic pressures measured by the Food and Drug Administrationapproved Omron HEM-790IT monitor using three different formulas given in the literature. The performance of our proposed method is compared with the maximum amplitude and zero-crossing methods in terms of mean error (ME), mean absolute error, and standard deviation of error (SDE). It is found that our proposed method achieves improvements of more than 20% in SDE compared with the maximum amplitude method and more than 50% in ME compared with the zero-crossing method.

Journal of Engineering and Technology, 2012

IEEE Transactions on Instrumentation and Measurement, 2014

2010 IEEE Instrumentation & Measurement Technology Conference Proceedings, 2010

We present a prototype of an integrated blood pressure (BP) and electrocardiogram (ECG) device fo... more We present a prototype of an integrated blood pressure (BP) and electrocardiogram (ECG) device for multi-parameter physiologic monitoring. A standard BP pressure cuff and an ordinary wristband have been modified to incorporate in them dry ECG electrodes made of thin conductive fabric. The modified BP cuff and wristband are coupled with commercially available hardware and software to harvest simultaneous arterial pulse wave and ECG data from the arm and wrist of the other hand. Software has been written for assessing multiple physiologic parameters from the harvested pulse wave and ECG signals. We provide an initial validation of the performance of our prototype by conducting a study on six healthy subjects.

2010 IEEE Instrumentation & Measurement Technology Conference Proceedings, 2010

ABSTRACT This paper addresses the suppression of the effects of the breathing signal from short d... more ABSTRACT This paper addresses the suppression of the effects of the breathing signal from short duration oscillometric waveform (OMW) recordings to obtain improved blood pressure estimates. As the amplitude modulating effects due to the breathing signal are multiplicative in nature, homomorphic filtering is done on the OMW. An adaptive filtering methodology is adopted to suppress the breathing signal from the OMW. For suppressing the breathing signal, an adaptive noise canceller (ANC) scheme is used when simultaneously acquired reference electrocardiogram (ECG) signal is available while an adaptive line enhancer (ALE) scheme is used when such a reference signal is not readily available. Existing algorithms are used for estimating the blood pressure values. After the suppression of the breathing effects, an improvement in the pressure estimates is observed. Unlike the current methodologies for suppressing the breathing effects from blood pressure measurements, the ALE scheme used in this paper does not require an additional reference signal.

2014 IEEE International Symposium on Medical Measurements and Applications (MeMeA), 2014

The final published version of this paper is available at: http://ieeexplore.ieee.org/ Please cit... more The final published version of this paper is available at: http://ieeexplore.ieee.org/ Please cite this article as follows: M. Forouzanfar, H.R. Dajani, V.Z. Groza, and M. Bolic, "Model-based oscillometric blood pressure estimation," IEEE Int.

2009 IEEE Intrumentation and Measurement Technology Conference, 2009

Three different algorithms for finding blood pressure through the oscillometric method were resea... more Three different algorithms for finding blood pressure through the oscillometric method were researched and assessed. It is shown that these algorithms are based on two different underlying approaches. The estimated values of systolic and diastolic blood pressure are compared against the nurse readings. The best two approaches turned out to be the linear approximation algorithm and the points of rapidly

2009 IEEE International Workshop on Medical Measurements and Applications, 2009

We propose a wavelet-based spectral density estimation method for characterizing pulse rate varia... more We propose a wavelet-based spectral density estimation method for characterizing pulse rate variability of short duration oscillometric blood pressure signals produced by a digital blood pressure monitor during routine measurements. To validate our wavelet metric we compare its performance with other techniques by studying correlations of pulse rate variability with age and mean arterial pressure. Our results indicate that the proposed wavelet metric offers a superior and accurate characterization of variability of short duration oscillometric blood pressure signals.

The Journal of the Acoustical Society of America, 2005

A filterbank-based algorithm for time-varying spectral analysis is proposed. The algorithm, which... more A filterbank-based algorithm for time-varying spectral analysis is proposed. The algorithm, which is an enhanced realization of the conventional spectrogram, consists of hundreds or thousands of highly overlapping wideband filter/detector stages, followed by a peak detector that probes the filter/detector outputs at very short time intervals. Analysis with synthetic modulated signals illustrates how the proposed method demodulates these signals. The resulting spectrogram-like display, referred to as a ''fine structure spectrogram,'' shows the fine structure of the modulations in substantially higher detail than is possible with conventional spectrograms. Error evaluation is performed as a function of various parameters of a single-and two-component synthetic modulated signal, and of parameters of the analysis system. In speech, the fine structure spectrogram can detect small frequency and amplitude modulations in the formants. It also appears to identify additional significant time-frequency components in speech that are not detected by other methods, making it potentially useful in speech processing applications.

IEEE journal of biomedical and health informatics, Jan 27, 2016

Non-invasive blood pressure (BP) measurement is an important tool for managing hypertension and c... more Non-invasive blood pressure (BP) measurement is an important tool for managing hypertension and cardiovascular disease. However, automated non-invasive BP measurement devices, which are usually based on the oscillometric method, do not always provide accurate estimation of BP. It has been found that change in arterial stiffness (AS) is an underlying mechanism of disagreement between an oscillometric BP monitor and a sphygmomanometer. This problem is addressed by incorporating parameters related to AS in the algorithm for BP measurement. Pulse transit time (PTT) is first used to estimate AS parameters, which are fixed into a model of the oscillometric envelope. This model can then be used to perform curve fitting to the measured signal using only four parameters: systolic BP, diastolic BP, mean BP, and lumen area at zero transmural pressure. The proposed technique is independent of the experimentally-determined characteristic ratios that are commonly used in existing oscillometric me...

IEEE International Workshop on Medical Measurement and Applications, 2010

MeMeA 2010 is the fifth edition of a successful international workshop series in the field of mea... more MeMeA 2010 is the fifth edition of a successful international workshop series in the field of measurements in medicine and health care, with previous editions held in: Cetraro-Italy (2009), Ottawa-Canada (2008), Warsaw-Poland (2007), and Benevento-Italy (2006).

Canadian Acoustics, Sep 1, 2007

institute o

Canadian Acoustics, Sep 1, 2009

A desirable feature of modern hearing aids is the ability to automatically adjust its behavior in... more A desirable feature of modern hearing aids is the ability to automatically adjust its behavior in different acoustic environments. There are two kinds of approaches that can achieve this. One is based on "classification" of environments, another is "direct learning" of preferred hearing aid settings. Our focus is on the second approach, in which an artificial neural network (ANN) learns the preferred volume setting of the hearing aid user. The performance of such a direct-learning system strongly depends on the chosen signal features. While a large number of features have been derived for environment classification , we are now aware of any that have been derived specifically for "direct learning", which has different requirements. For example, environment classification should not in general be sensitive to the sound volume, whereas direct learning of volume setting not only depends on the volume, but also depends on how this volume affects speech intelligibility and user comfort. Moreover, the preferred volume setting depends on the hearing loss profile, and whether it is profound, severe, or moderate. The goal of this work is to derive suitable features that the ANN will use to set the volume such that it optimizes speech intelligibility. New features are proposed, which are based on measures of speech intelligibility, namely the Speech Intelligibility Index (SII) (ANSI S3.5-1997) and the Coherence SII (CSII) . The performance of these features is then investigated using a simulator of a hearing aid user (SPOT, 2009).

2008 IEEE International Workshop on Medical Measurements and Applications, 2008

ABSTRACT

4th International Workshop on Soft Computing Applications, 2010

Feed-Forward Neural Network (FFNN) has recently been utilized to estimate blood pressure (BP) fro... more Feed-Forward Neural Network (FFNN) has recently been utilized to estimate blood pressure (BP) from the oscillometric measurements. However, there has been no study till now that consolidated the role played by the different neural network (NN) training algorithms in affecting the BP estimates. This paper compares the estimation errors in the BP due to ten different training algorithms belonging to three classes: steepest descent (with variable learning rate, with variable learning rate and momentum, resilient backpropagation), quasi-Newton (Broyden-Fletcher-Goldfarb-Shanno, one step secant, Levenberg-Marquardt) and conjugate gradient (Fletcher-Reeves update, Polak-Ribiére update, Powell-Beale restart, scaled conjugate gradient) that are used to train two separate NNs: one to estimate the systolic pressure and the other one to estimate the diastolic pressure. The different training algorithms are compared in terms of estimation error (mean absolute error and standard deviation of error) and training performance (training time and number of training iterations to reach the optimal weights). The NN-based approach is also compared with the conventional maximum amplitude algorithm.

Pitch discrimination experiments have demonstrated that human listeners can detect very small fre... more Pitch discrimination experiments have demonstrated that human listeners can detect very small frequency changes in stimuli of short duration. Inspired by this ability, an algorithm for high resolution time-varying spectral analysis is proposed. Mathematical analysis, with various types of synthetic modulated signals, demonstrates that the proposed method correctly demodulates these signals. The resulting spectrogramlike display, referred to as a 'Fine Structure Spectrogram', shows the fine structure of the modulations in higher detail than is possible with conventional spectrograms. With recorded speech samples, the fine structure spectrogram detects small frequency and amplitude modulations in the formants of speech. It also appears to identify additional components in speech that are not detected by other methods.

The American journal of physiology

The purpose of this study was to characterize oscillations in muscle sympathetic nerve activity (... more The purpose of this study was to characterize oscillations in muscle sympathetic nerve activity (MSNA) in the frequency domain in healthy subjects and patients with congestive heart failure (CHF) and to relate these to blood pressure (BP), heart rate (HR), and breathing frequency. MSNA burst frequency was significantly greater in CHF [52 +/- 21 (n = 12) vs. 35 +/- 11 (n = 19) bursts/min, P < 0.05], whereas breathing frequency and HR were similar. There was no significant difference between CHF and healthy subjects in total power, harmonic power, and nonharmonic power in the MSNA spectrum from 0 to 0.5 Hz, but low frequency power (LF, 0.05-0.15 Hz, P < 0.05) was reduced in heart failure patients. There was less coherence between BP and MSNA in the LF range, but similar spectral power in both groups in the very LF (VLF, 0-0.05 Hz) and high frequency (0.15-0.5 Hz) ranges. The transfer of MSNA oscillations into BP in the VLF (P < 0.05) and LF (P < 0.02) ranges was significantly lower in CHF, but gains in the transfer function and in the coherence between BP and MSNA and in the coherence between respiration and MSNA were similar in the two groups. These observations indicate that modulation of MSNA by the arterial baroreflex and respiration is preserved in CHF. The loss of LF power in the MSNA signal may be due to impaired neuroeffector transduction. The higher sympathetic nerve firing rate in CHF would therefore appear to be caused by factors other than the loss of regulation by these two inhibitory influences.

The Journal of the Acoustical Society of America, 2015

2015 IEEE International Symposium on Medical Measurements and Applications (MeMeA) Proceedings, 2015

IEEE Transactions on Instrumentation and Measurement, 2015

An accurate noninvasive estimation of mean arterial pressure (MAP) is of great importance in the ... more An accurate noninvasive estimation of mean arterial pressure (MAP) is of great importance in the evaluation of circulatory function and prognosis of some cardiovascular diseases. This paper proposes a novel oscillometric MAP estimation method based on the dependence of pulse transit time (PTT) on cuff pressure (CP). The PTT computed as the time interval between the electrocardiogram (ECG) R-peaks and the maximum slope points on the oscillometric pulses is mathematically modeled by considering the cuff-arm-artery system and the blood flow dynamics. It is then analytically shown that MAP can be approximated as the CP at which the PTT is maximum. Based on our theoretical findings, a new method of MAP estimation from simultaneous ECG and oscillometric blood pressure measurements is proposed. Our proposed method is validated with a pilot study in which 150 recordings from 10 subjects are analyzed. The reference MAP is computed from the systolic and diastolic pressures measured by the Food and Drug Administrationapproved Omron HEM-790IT monitor using three different formulas given in the literature. The performance of our proposed method is compared with the maximum amplitude and zero-crossing methods in terms of mean error (ME), mean absolute error, and standard deviation of error (SDE). It is found that our proposed method achieves improvements of more than 20% in SDE compared with the maximum amplitude method and more than 50% in ME compared with the zero-crossing method.

Journal of Engineering and Technology, 2012

IEEE Transactions on Instrumentation and Measurement, 2014

2010 IEEE Instrumentation & Measurement Technology Conference Proceedings, 2010

We present a prototype of an integrated blood pressure (BP) and electrocardiogram (ECG) device fo... more We present a prototype of an integrated blood pressure (BP) and electrocardiogram (ECG) device for multi-parameter physiologic monitoring. A standard BP pressure cuff and an ordinary wristband have been modified to incorporate in them dry ECG electrodes made of thin conductive fabric. The modified BP cuff and wristband are coupled with commercially available hardware and software to harvest simultaneous arterial pulse wave and ECG data from the arm and wrist of the other hand. Software has been written for assessing multiple physiologic parameters from the harvested pulse wave and ECG signals. We provide an initial validation of the performance of our prototype by conducting a study on six healthy subjects.

2010 IEEE Instrumentation & Measurement Technology Conference Proceedings, 2010

ABSTRACT This paper addresses the suppression of the effects of the breathing signal from short d... more ABSTRACT This paper addresses the suppression of the effects of the breathing signal from short duration oscillometric waveform (OMW) recordings to obtain improved blood pressure estimates. As the amplitude modulating effects due to the breathing signal are multiplicative in nature, homomorphic filtering is done on the OMW. An adaptive filtering methodology is adopted to suppress the breathing signal from the OMW. For suppressing the breathing signal, an adaptive noise canceller (ANC) scheme is used when simultaneously acquired reference electrocardiogram (ECG) signal is available while an adaptive line enhancer (ALE) scheme is used when such a reference signal is not readily available. Existing algorithms are used for estimating the blood pressure values. After the suppression of the breathing effects, an improvement in the pressure estimates is observed. Unlike the current methodologies for suppressing the breathing effects from blood pressure measurements, the ALE scheme used in this paper does not require an additional reference signal.

2014 IEEE International Symposium on Medical Measurements and Applications (MeMeA), 2014

The final published version of this paper is available at: http://ieeexplore.ieee.org/ Please cit... more The final published version of this paper is available at: http://ieeexplore.ieee.org/ Please cite this article as follows: M. Forouzanfar, H.R. Dajani, V.Z. Groza, and M. Bolic, "Model-based oscillometric blood pressure estimation," IEEE Int.

2009 IEEE Intrumentation and Measurement Technology Conference, 2009

Three different algorithms for finding blood pressure through the oscillometric method were resea... more Three different algorithms for finding blood pressure through the oscillometric method were researched and assessed. It is shown that these algorithms are based on two different underlying approaches. The estimated values of systolic and diastolic blood pressure are compared against the nurse readings. The best two approaches turned out to be the linear approximation algorithm and the points of rapidly

2009 IEEE International Workshop on Medical Measurements and Applications, 2009

We propose a wavelet-based spectral density estimation method for characterizing pulse rate varia... more We propose a wavelet-based spectral density estimation method for characterizing pulse rate variability of short duration oscillometric blood pressure signals produced by a digital blood pressure monitor during routine measurements. To validate our wavelet metric we compare its performance with other techniques by studying correlations of pulse rate variability with age and mean arterial pressure. Our results indicate that the proposed wavelet metric offers a superior and accurate characterization of variability of short duration oscillometric blood pressure signals.

The Journal of the Acoustical Society of America, 2005

A filterbank-based algorithm for time-varying spectral analysis is proposed. The algorithm, which... more A filterbank-based algorithm for time-varying spectral analysis is proposed. The algorithm, which is an enhanced realization of the conventional spectrogram, consists of hundreds or thousands of highly overlapping wideband filter/detector stages, followed by a peak detector that probes the filter/detector outputs at very short time intervals. Analysis with synthetic modulated signals illustrates how the proposed method demodulates these signals. The resulting spectrogram-like display, referred to as a ''fine structure spectrogram,'' shows the fine structure of the modulations in substantially higher detail than is possible with conventional spectrograms. Error evaluation is performed as a function of various parameters of a single-and two-component synthetic modulated signal, and of parameters of the analysis system. In speech, the fine structure spectrogram can detect small frequency and amplitude modulations in the formants. It also appears to identify additional significant time-frequency components in speech that are not detected by other methods, making it potentially useful in speech processing applications.