Abstract: Coherence analysis is widely employed to study the correlation between discharge times of simultaneously active motor units. Despite the widespread application of the technique, it has not been fully established how the characteristics of the observed coherence are related to the properties of the shared motoneuron inputs. In addition, the exact relationship between coherence and traditional measures of motor unit synchronization remains unclear. The purpose of this study was to examine the influence of shared motoneuron inputs on coherence between motor unit discharge patterns using computer simulations. Although less sensitive to motor unit firing rates than traditional synchronization-based indices, coherence tended to decrease with increasing frequency of the common input and to increase slightly when the common input frequency was close to the motor unit firing rates. In addition, coherence tended to be highest between motor units with similar firing rates. A linear association was observed between synchronization and coherence in the 15-30 Hz range and between 'common drive' and coherence in the 0-5 Hz range. The results suggest that caution should be taken when interpreting differences in coherence observed between motor units with significantly different firing properties or when comparing data with coherence in different frequency ranges.
Abstract: Ambulatory respiratory data was gathered using inductive lethysmography technology with synchronous ECG(LifeShirte , VivoMetrics, Ventura, CA) during a study to evaluate the effect of an anxiolytic on heart rate variability and respiratory pattern as indicators of anxiety state. Positive control (PCR; post-marketing, broadly prescribed anxiolytic)and placebo (PBO) data was included in the analysis. Tidal volume waveforms were the result of a weighted sum of the abdominal and rib cage IP bands according to the qualitative diagnostic calibration method. A breath detection algorithm was run to identify the beginning and end of inhalation in these waveforms. Several types of respiratory artifact are common with ambulatory, non-controlled recordings and a consistent and reliable means is necessary to identify and manage such artifacts. An automated approach was adopted to define a reliable breathing index for each breath that labels that breath as contaminated by artifact or not. The root mean square of successive differences (RMSSD) were computed on the tidal inspiratory volumes and total breath times for each epoch, both for all breaths and for only those breaths that were labeled as reliable. The results indicate that when a priori automated artifact detection is included, there is a significant linear decrease in both the volume and time indices for the PCR, whilst no significant differences were noted in the PBO group. Analyzing the data without prior marking of reliable breaths showed no significant results for either group. This study demonstrates the validity of ambulatory respiratory measurements as a means to assess anxiety and establishes the need to first identify reliable breathing periods prior to the analysis of ambulatory respiratory data.
Abstract: MOTIVATION: Genotype-phenotype modeling problems are often overcomplete, or ill-posed, since the number of potential predictors-genes, proteins, mutations and their interactions-is large relative to the number of measured outcomes. Such datasets can still be used to train sparse parameter models that generalize accurately, by exerting a principle similar to Occam's Razor: When many possible theories can explain the observations, the most simple is most likely to be correct. We apply this philosophy to modeling the drug response of Type-1 Human Immunodeficiency Virus (HIV-1). Owing to the decreasing expense of genetic sequencing relative to in vitro phenotype testing, a statistical model that reliably predicts viral drug response from genetic data is an important tool in the selection of antiretroviral therapy (ART). The optimization techniques described will have application to many genotype-phenotype modeling problems for the purpose of enhancing clinical decisions. RESULTS: We describe two regression techniques for predicting viral phenotype in response to ART from genetic sequence data. Both techniques employ convex optimization for the continuous subset selection of a sparse set of model parameters. The first technique, the least absolute shrinkage and selection operator, uses the l(1) norm loss function to create a sparse linear model; the second, the support vector machine with radial basis kernel functions, uses the epsilon-insensitive loss function to create a sparse non-linear model. The techniques are applied to predict the response of the HIV-1 virus to 10 reverse transcriptase inhibitor and 7 protease inhibitor drugs. The genetic data are derived from the HIV coding sequences for the reverse transcriptase and protease enzymes. When tested by cross-validation with actual laboratory measurements, these models predict drug response phenotype more accurately than models previously discussed in the literature, and other canonical techniques described here. Key features of the methods that enable this performance are the tendency to generate simple models where many of the parameters are zero, and the convexity of the cost function, which assures that we can find model parameters to globally minimize the cost function for a particular training dataset. AVAILABILITY: Results, tables and figures are available at ftp://ftp.genesecurity.net. SUPPLEMENTARY INFORMATION: An Appendix to accompany this article is available at Bioinformatics online.
Abstract: We describe the use of the l1 norm for selection of a sparse set of model parameters that are used in the prediction of viral drug response, based on genetic sequence data of the Human Immunodeficiency Virus (HIV) reverse-transcriptase enzyme. We discuss the use of the l1 norm in the Least Absolute Selection and Shrinkage Operator (LASSO) regression model and the Support Vector Machine model. When tested by cross-validation with laboratory measurements, these models predict viral phenotype, or resistance, in response to Reverse-Transcriptase Inhibitors (RTIs) more accurately than other known models. The l1 norm is the most selective convex function, which sets a large proportion of the parameters to zero and also assures that a single optimal solution will be found, given a particular model formulation and training data set. A statistical model that reliably predicts viral drug response is an important tool in the selection of Anti-Retroviral Therapy. These techniques have general application to modeling phenotype from complex genetic data.
Abstract: Using a number of different homeostatic control mechanisms in the brain and peripheral physiological systems, metabolic activity is continuously regulated at rest and during exercise to prevent catastrophic system failure. Essential for the function of these regulatory processes are baseline "setpoint" levels of metabolic function, which can be used to calculate the level of response required for the maintenance of system homeostasis after system perturbation, and to which the perturbed metabolic activity levels are returned to at the end of the regulatory process. How these setpoint levels of all the different metabolic variables in the different peripheral physiological systems are created and maintained, and why they are similar in different individuals, has not been well explained. In this article, putative system regulators of metabolic setpoint levels are described. These include that: (i) innate setpoint values are stored in a certain region of the central nervous system, such as the hypothalamus; (ii) setpoint values are created and maintained as a response to continuous external perturbations, such as gravity or "zeitgebers", (iii) setpoint values are created and maintained by complex system dynamical activity in the different peripheral systems, where setpoint levels are regulated by the ongoing feedback control activity between different peripheral variables; (iv) human anatomical and biomechanical constraints contribute to the creation and maintenance of metabolic setpoints values; or (v) a combination of all these four different mechanisms occurs. Exercise training and disease processes can affect these metabolic setpoint values, but the setpoint values are returned to pre-training or pre-disease levels if the training stimulus is removed or if the disease process is cured. Further work is required to determine what the ultimate system regulator of metabolic setpoint values is, why some setpoint values are more stringently protected by homeostatic regulatory mechanisms than others, and the role of conscious decision making processes in determining the regulation of metabolic setpoint values.
Abstract: The functional coupling between the primary motor cortex (M1) and the supplementary motor area (SMA) in the generation of internally paced versus externally cued rhythmic movements was explored using electroencephalography (EEG). This has important implications for the study of Parkinsonian patients who demonstrate decreased ability to perform internally paced rhythmic movement tasks. In particular, the temporal evolution of the coherence between M1 and SMA was studied using a recently developed time-frequency wavelet coherence algorithm. As this approach is not reliant upon pooling data from multiple trials to form a single estimate of the coherence for each subject, a subject by subject comparison is possible to determine subject specific frequencies of interest. It was found that at certain frequencies, the coupling between M1 and SMA was increased in the internally paced versus the externally triggered movement task. At only these specific frequencies did the peak of the coherence in the internal task precede that of the peak in the external task, which was time locked with the movement onset. This suggests a dual role of the M1-SMA coupling for both movement preparation and movement execution.
Abstract: BACKGROUND: In investigations of the human motor system, two approaches are generally employed toward the identification of common modulating drives from motor unit recordings. One is a frequency domain method and uses the coherence function to determine the degree of linear correlation between each frequency component of the signals. The other is a time domain method that has been developed to determine the strength of low frequency common modulations between motor unit spike trains, often referred to in the literature as 'common drive'. METHODS: The relationships between these methods are systematically explored using both mathematical and experimental procedures. A mathematical derivation is presented that shows the theoretical relationship between both time and frequency domain techniques. Multiple recordings from concurrent activities of pairs of motor units are studied and linear regressions are performed between time and frequency domain estimates (for different time domain window sizes) to assess their equivalence. RESULTS: Analytically, it may be demonstrated that under the theoretical condition of a narrowband point frequency, the two relations are equivalent. However practical situations deviate from this ideal condition. The correlation between the two techniques varies with time domain moving average window length and for window lengths of 200 ms, 400 ms and 800 ms, the r2 regression statistics (p < 0.05) are 0.56, 0.81 and 0.80 respectively. CONCLUSIONS: Although theoretically equivalent and experimentally well correlated there are a number of minor discrepancies between the two techniques that are explored. The time domain technique is preferred for short data segments and is better able to quantify the strength of a broad band drive into a single index. The frequency domain measures are more encompassing, providing a complete description of all oscillatory inputs and are better suited to quantifying narrow ranges of descending input into a single index. In general the physiological question at hand should dictate which technique is best suited.
Abstract: The dimensions and wall elasticity commonly taper along the length of mammalian arteries. The effects of taper on flow velocity waveforms can be included by either of two methods; to theoretically divide the artery into short sections wherein the properties are assumed constant (the approximate solution); or to find an exact solution incorporating the effects of taper. In this paper, an exact solution to the resulting, and previously unsolved nonlinear Ricatti equation for the impedance, is obtained by a process of substitutions. This solution is utilised to develop an exact expression for the flow velocity in the artery. The transmission line equations are then combined into a single integral expression for the entire artery and an exact solution to this is evaluated. This is the first solution to simultaneously account for both geometric and elastic taper, and it has been validated by comparing simulations of flow in the aorta of a dog to those using an infinitesimal approximate solution. The Pulsatility Index of the approximate solution requires at least 10 segments to converge to within 5% of that using the exact solution. The exact solution thus accurately accounts for the effects of exponential taper, and may be used to improve existing arterial models, which use the less accurate and more computationally cumbersome approximate solution.
Abstract: Rectification of the electromyographic (EMG) signal is a commonly used pre-processing procedure that allows detection of significant coherence between EMG and measured cortical signals. However, despite its accepted and wide-spread use, no detailed analysis has been presented to offer insight into the precise function of rectification. We begin this paper with arguments based on single motor unit action potential (AP) trains to demonstrate that rectification effectively enhances the firing rate information of the signal. Enhancement is achieved by shifting the peak of the AP spectrum toward the lower firing rate frequencies, whilst maintaining the firing rate spectra. A similar result is obtained using the analytic envelope of the signal extracted using the Hilbert transform. This argument is extended to simulated EMG signals generated using a published EMG model. Detection of firing rate frequencies is obtained using phase randomised surrogate data, where the original EMG power spectrum exceeds the averaged rectified surrogate spectra at integer multiples of firing rate frequencies. Model simulations demonstrate that this technique accurately determines grouped firing rate frequencies. Extraction of grouped firing rate frequencies prior to coherency analyses may further aid interpretation of significant cortico-muscular coherence findings.
Abstract: The circulation of a human fetus has been modeled using a transfer function that is based on the arterial dimensions at 28 weeks gestational age (GA). These dimensions have then been adapted for growth between 28 and 40 weeks GA. The input to the model is a series of current pulses at the fetal heart rate, where current in the model is analogous to volume blood flow in the fetus. The arterial system is divided into short segments that are cascaded together. The respective transfer functions are based on the dimensions, wall properties and fluid characteristics at each frequency and GA. Bleed off conductances distribute current to circuits representing the various anatomical regions. In particular, the placenta is simplified to a symmetrically distributed network of branching vessels, each represented by a transfer function. All calculations are performed in the frequency domain, after which the inverse Fourier transform is used to calculate the currents that represent the time-domain blood flow waveforms. Simulated flow waveform resistance index and pulsatility index values are within 8% of those reported for human clinical studies, at all gestational ages.
Abstract: A transmission line model of the human foetal circulatory system is presented. The model has been developed in the frequency domain with the cardiac input modeled as a flow rather than as a pressure pulse and is structured upon electrical transmission line analogies. The model is formed by cascading solutions to the two-dimensional Navier-Stokes equations for both oscillatory and steady, laminar viscous fluid flow in isotropic visco-elastic tubes with thick walls, which are constrained by surrounding tissues. Simulations allow for representation of both forward and retrograde travelling flow and pressure waves in all of the main foetal arterial vessels. The solution is verified by a comparison of model generated Doppler indices in the thoracic aorta, abdominal aorta, iliac artery and both ends of the umbilical arteries with previously published indices obtained by clinical measurements in these arteries. For simulations of blood flow in a healthy foetus, the model generated Pulsatility and Resistance indices were on average within 8% of the corresponding clinical measurements. The model results also demonstrates that placental resistance must increase by a factor of three, corresponding to a 60% decrease in flow to the placenta, before umbilical arterial absent end diastolic flow is observed. Differences between indices obtained from simulations at opposite ends of the umbilical arteries increase with increasing placental resistance.
Abstract: An analytical solution for pulsatile axial flow velocity waveforms in curved elastic tubes is presented. The result is obtained by exact solution of linearized Navier-Stokes and tube motion equations in a torroidal coordinate system. Fourier analysis is used to divide the flow into constant and oscillatory components which are separately considered. The solution is used to investigate the effects of curvature on volumetric axial velocity flow waveforms, as would be measured by Doppler ultrasound techniques. In typical human arteries, the greatest effects of curvature on the volumetric axial flow are exerted on the constant component and at low values of the frequency parameter for the oscillatory components. Here, the magnitude and phase angle of oscillatory flow in the curved tube, relative to that in the straight tube, differ by maximum values of 1.2% and 0.15 rad, respectively. However, constant flow may vary by as much as 60% at high Dean numbers. The solution is presented in a form similar to Womersley's solution for the straight elastic tube and may, thus, be incorporated into a transmission-line analog model. These models are frequently used to investigate axial flow velocity variations in mamillian circulatory systems and this work offers a tool which may extend these models to incorporate the effects of curvature.
Abstract: The development of a robust algorithm allowing good recognition of low-quality fingerprints with inexpensive hardware is investigated. A threshold FFT approach is developed to simultaneously smooth and enhance poor quality images derived from a database of imperfect prints. Features are extracted from the enhanced images using a number of approaches including a novel wedge ring overlay minutia detector that is particularly robust to imperfections. Finally, a number of neural net and statistically based classifiers are evaluated for the recognition task. Results for various combinations of the process are presented and discussed with regard to their utility in such a system.
Abstract: The objective of this study was to identify a parsimonious set of physiological measures that could be used to best predict cognitive fatigue levels. A 37 hour sleep deprivation study was conducted to induce reduced levels of alertness and cognitive impairment as measured by a psychomotor vigilance test. Non-invasive, wearable and ambulatory sensors were used to acquire cardio-respiratory and motion data during the sleep deprivation. Subsequently 23 potential predictors were derived from the raw sensor data. The least absolute shrinkage and selection operator, along with a cross validation strategy was used to create a sparse model and identify a minimum predictor subset that provided the best prediction accuracy. Final predictor selection was found to vary with task and context. Depending on context selected predictors indicated elevated levels of sympathetic nervous system activity, increased restlessness during engaging tasks and increased cardio-respiratory synchronization with increasing cognitive fatigue.
Abstract: The seven papers in this slim volume apply mathematical modeling techniques to study biological fluid-wall systems, particularly the human cardiovascular system. The two longest papers use one-dimensional nonlinear systems to model blood pulse propagation in compliant arteries, and a simplified physical model to explore the effects of weak wall permeability on lung airway reopening. The other topics are the elastodynamics of saccular aneurysms, a training device for skeletal muscle in cardiac assist use, umbilical cord coiling and ductus venosus dilation, an integral formulation for fluid-structure interaction in hemodynamics, and a differential model of blood flow in a stented artery.
Abstract: Objectives: The objective of this study was to identify a sparse set of physiological measures that could
be used to predict cognitive fatigue levels. PVT mean reaction time reaction time and lapses were used as
a ‘ground truth’ measure of performance impairment indicative of reduced alertness.
Approach and methods: Six subjects were recruited for a continuous 37 hour sleep deprivation study.
Subject physiologic data were recorded using commercially available, wearable, ambulatory
physiological monitoring systems. These systems allow unbiased, non-invasive and continuous
recordings of cardio-respiratory and activity measures. Subjects further performed the PVT every hour
for the full study duration.
A suitable set of attributes that have relevant physiological meaning and compactly represent the original
physiological data set was identified. This was based upon the demonstrated fact that there is a strong
link between cognitive fatigue and autonomic nervous system activity. In particular, the fatigue state is
associated with a shift of sympathovagal balance toward sympathetic predominance and reduced vagal
tone. In addition, respiratory instability has been demonstrated to be a strong predictor of
psychophysiologic state. Initial feature selection focused on the inclusion of cardio-respiratory markers
that could be used to infer autonomic activation, resulting in a large set of possible predictive features.
These included indices of: heart rate, heart rate variability (HRV- several indices), tidal volume, breathing
rate, respiratory instability (RI - several indices), motion, postural shifts and skin temperature. Features
were extracted over a 3 minute quiet period preceding, during and following the PVT test.
When the number of predictors exceeds the number of training samples, the modeling problem is
underdetermined, or ill-posed in the Hadamard sense. In this instance, it is desirable to find a model with
significantly fewer predictors and in fact, the more sparse the model, the more likely that the predictors
are causally related to the dependent variable. In order to achieve this for ill-posed data, the values of the
regression coefficients can be constrained via a shrinkage function. We used the LASSO shrinkage and
selection method, along with the least angle regression method for solving the LASSO, to create a sparse
model and thus determine an optimal feature subset. The resultant subset should therefore consist of the
most relevant predictors of cognitive fatigue.
Results: The LASSO only selected two variables: the normalized low frequency content of the HRV
(LFnorm) and the tidal volume instability (TVI), which is an index of RI. These variables both indicate an
increase in sympathetic arousal associated with attention toward a task.
Conclusion: The LASSO technique allows one to select en-masse, via a continuous subset optimization, a
set of cardio-respiratory variables that together are effective predictors of operator alertness status. This
technique combined with commercially-available, wearable physiologic monitoring systems will result in
a system that can improve operational safety and effectiveness by accurately assessing cognitive fatigue
levels during stressful day to day conditions.
Abstract: Objective: We investigated whether a recently introduced 'complexity' metric, Sample Entropy (Lake et al Comp Physiol 2002), quantifies periodic breathing during exercise and hypoxia.
Methods: We evaluated 6 subjects at rest and during walking at 3610m and 4559m while climbing Mt. Rosa. A portable inductive plethysmograph (LifeShirt, VivoMetrics, CA) recorded breathing patterns and minute ventilation (V'E), resampled to1Hz. Normalized sample entropy was computed for V'E; lower values imply more structure in a signal. Fast Fourier Transforms of V'E time series provided normalized peak power which increases with increasing periodic breathing.
Results: Periodic breathing was hardly discernible by eye in V'E tracings. Peak power and sample entropy changed in opposite directions with transition to walking (P<0.05) and at higher altitude (NS).
Conclusions: Complexity metrics are sensitive means to quantify effects of physical activity on control of breathing at high altitude. Sample entropy incorporates correlations over multiple time scales and is not limited to stationary signals so that it can be applied during exercise. Unlike power spectral analysis, sample entropy quantifies breathing regularity by a single value which is advantageous if comparisons are made during changes in frequency of periodic breathing.
Normalized Sample Entropy Normalized Peak Power
3610m 4559m 3610m 4559m
Rest 0.277±0.033 0.243±0.074 0.086±0.015 0.108±0.027
Walking* 0.564±0.084 0.498±0.133 0.029±0.006 0.037±0.008
Means ±SE; * P<0.05 ANOVA for effect of walking for both analysis methods
Abstract: Rectification of the electromyographic (EMG) signal is a commonly used pre-processing procedure that allows detection of significant coherence between EMG and measured cortical signals. It has been demonstrated that rectification enhances mean group firing rates. This paper develops an efficient algorithm that may be used to extract these relevant frequencies. This method assumes the EMG to be a realization of a randomly distributed stationary process. Therefore the 'true' power spectrum of this process will not contain specific timing information. However the power spectrum of the actual rectified EMG does contain timing information which is detectable at frequencies where it exceeds the analytic power spectrum. Coherence of measured electroencephalographic (EEG) signals and rectified EMG show that significant coherence peaks occur at the same frequencies obtained using the algorithm. Cortical drive appears to synchronize motor unit firings to particular frequencies that may be extracted from the rectified EMG alone. These frequencies are not easily obtained from either the EEG or the non-rectified EMG.
Abstract: A number of descending drives tend to synchronise motor unit firings to particular frequencies during sustained voluntary contractions. Detection of these drives is usually performed using bivariate or multivariate measurements. We investigate the possibility of detecting these drives from single motor unit recordings. A computational model of the surface electromyogram was used to examine the variability of the instantaneous firing rate intervals of synchronised motor unit trains. The model generated distributions of this variability suggest that they arise from a mixture of two separate Gaussian distributions. A constrained version of the expectation-maximization algorithm is used to determine the component density means which are related to the descending drive frequencies. This algorithm is tested against a variety of drive frequencies and drive strengths.
