Automated correction of unwanted phase jumps in reference signals which corrupt MRSI spectra after eddy current correction.

A commonly applied step in the postprocessing of gradient localized proton MR spectroscopy, is correction for eddy current effects using the water signal as a reference. However, this method can degrade some of the metabolite signals, in particular if applied on proton MR spectroscopic imaging data. This artifact arises from the water reference signal in the presence of a second signal which resonates close to the main water resonance. The interference of both resonances will introduce jumps in the phase of the reference time domain signal. Using this phase for eddy current correction will result in a ringing artifact in the frequency domain of the metabolite signal over the whole frequency range. We propose a moving window correction algorithm, which screens the phase of reference signals and removes phase jumps in time domain caused by interference of signals from multiple spin systems. The phase jumps may be abrupt or gradually distributed over several time data points. Because the correction algorithm only corrects time data points which contain phase jumps, the phase is minimally disrupted. Furthermore, the algorithm is automated for large datasets, correcting only those water reference signals which are corrupted. After correction of the corrupted reference signals, normal eddy current correction may be performed. The algorithm is compared with a method which uses a low-pass filter and tested on simulated data as well as on in vivo proton spectroscopic imaging data from a healthy volunteer and from patients with a brain tumor.

Automatic correction for phase shifts, frequency shifts, and lineshape distortions across a series of single resonance lines in large spectral data sets.

A new model-free method is presented that automatically corrects for phase shifts, frequency shifts, and additional lineshape distortions of one single resonance peak across a series of in vivo NMR spectra. All separate phase and frequency variations are quickly and directly derived from the common lineshape in the data set using principal component analysis and linear regression. First, the new approach is evaluated on simulated data in order to quantitatively assess the phase and frequency shifts which can be removed by the proposed correction procedure. Subsequently, the value of the method is demonstrated on in vivo (31)P NMR spectra from skeletal muscle of the hind leg of the mouse focusing on the phosphocreatine resonance which is distorted by the experimental procedure. Phase shifts, frequency shifts, and lineshape distortions with respect to the common lineshape in the spectral data set could successfully be removed.

Selectivity for temporal characteristics of sound and interaural time difference of auditory midbrain neurons in the grassfrog: a system theoretical approach.

The selectivity for temporal characteristics of sound and interaural time difference (ITD) was investigated in the torus semicircularis (TS) of the grassfrog. Stimuli were delivered by means of a closed sound system and consisted of binaurally presented Poisson distributed condensation clicks, and pseudo-random (RAN) or equidistant (EQU) click trains of which ITD was varied. With RAN and EQU trains, 86% of the TS units demonstrated a clear selectivity for ITD. Most commonly, these units had monotonically increasing ITD-rate functions. In general, units responding to Poisson clicks, responded also to RAN and EQU trains. One category of units which showed strong time-locking had comparable selectivities for ITD with both stimulus ensembles. A second category of units showed a combined selectivity for temporal structure and ITD. These units responded exclusively to EQU trains in a nonsynchronized way. From the responses obtained with the Poisson click ensemble so-called Poisson system kernels were determined, in analogy to the Wiener-Volterra functional expansion for nonlinear systems. The kernel analysis was performed up to second order. Contralateral (CL) first order kernels usually had positive or combinations of positive and negative regions, indicating that the contralateral ear exerted an excitatory or combined excitatory-inhibitory influence upon the neural response. Ipsilateral (IL), units were characterized by first order kernels which were not significantly different from zero, or kernels in which a single negative region was present. A large variety of CL second order kernels has been observed whereas rarely IL second order kernels were encountered. About 35% of the units possessed nonzero second order cross kernels, which indicates that CL and IL neural processes are interacting in a nonlinear way. Units demonstrating a pronounced selectivity for ITD, were generally characterized by positive CL combined with negative IL first order kernels. Findings suggested that, in the grassfrog, neural selectivity for ITD mainly is established by linear interaction of excitatory and inhibitory processes originating from the CL and IL ear, respectively. Units exhibiting strong time-locking to Poisson clicks and RAN and EQU trains had significantly shorter response latencies than moderately time-locking units. In the first category of units, a substantial higher number of nonzero first and second order kernels was observed. It was concluded that nonlinear response properties, as observed in TS units, most likely have to be ascribed to nonlinear characteristics of neural components located in the auditory nervous system peripheral to the torus semicircularis.

A combined sensitivity for frequency and interaural intensity difference in neurons in the auditory midbrain of the grassfrog.

The relation between spectral tuning and sensitivity for interaural intensity difference (IID) was studied for single units in the auditory midbrain of the grassfrog. The stimuli consisted of sequences of pure tones of different frequency and interaural intensity differences presented by means of a closed sound system. At best excitatory frequency, three types of binaural interaction were observed: E0 (one ear excitatory 23%), EE (both ears excitatory 9%) and EI (one ear excitatory, the other inhibitory 67%). For a considerable number of units different types of binaural interaction were observed for different stimulus frequencies. More than 30% of the binaural units had multiple excitatory and inhibitory regions in their spectrotemporal selectivity. E0 and EI units had uniformly distributed best frequencies, EE units generally had best frequencies near 1.0 kHz. The E0 and EE categories had response latencies less than about 70 ms whereas EI units could have longer latencies. Most EE and all EI category units had sigmoidally shaped IID-rate curves. About 40% of the units had a combined sensitivity for sound spectrum and IID which was invariant to overall stimulus intensity. For nearly all EI units the inhibitory influence of the ipsilateral ear was confined to frequencies in the 0.4-1.6 kHz range and was not correlated with a unit's best frequency. By means of a simple additive model we demonstrated that determination of sound source laterality can be achieved by ensemble coding in the auditory midbrain.

Measuring and modelling the response of auditory midbrain neurons in the grassfrog to temporally structured binaural stimuli.

The combined selectivity for amplitude modulation frequency (AMF) and interaural time difference (ITD) was investigated for single units in the auditory midbrain of the grassfrog. Stimuli were presented by means of a closed sound system. A large number of units was found to be selective for AMF (95%) or ITD (85%) and mostly, these selectivities were intricately coupled. At zero ITD most units showed a band-pass (54%) or bimodal (24%) AMF-rate histogram. At an AMF of 36 Hz, which is equal to the pulse repetition rate of the mating call, 70% of the units possessed an asymmetrical ITD-rate histogram, whereas about 15% showed a symmetrically peaked histogram. With binaural stimulation more units appeared to be selective for AMF (95%) as was the case with monaural stimulation (85%). A large fraction of the units appeared to be most selective for ITD at AMFs of 36 and 72 Hz, whereas units seldomly exhibited ITD selectivity with unmodulated tones. Based upon previous papers (Melssen et al., 1990; Van Stokkum, 1990) a binaural model is proposed to explain these findings. An auditory midbrain neuron is modelled as a third order neuron which receives excitatory input from second order neurons. Furthermore the model neuron receives inputs from the other ear, which may be either excitatory or inhibitory. Spatiotemporal integration of inputs from both ears, followed by action potential generation, produces a combined selectivity for AMF and ITD. In particular the responses of an experimentally observed EI neuron to a set of stimuli are reproduced well by the model.

Sensitivity for interaural time and intensity difference of auditory midbrain neurons in the grassfrog.

The sensitivity for interaural time (ITD) and intensity (IID) difference was investigated for single units in the auditory midbrain of the grassfrog. A temporally structured stimulus was used which was presented by means of a closed sound system. At best frequency (BF) the majority of units was selective for ITD as indicated by an asymmetrically (73%) or symmetrically (7%) shaped ITD-rate histogram. About 20% appeared to be nonselective. Units with a symmetrical rate histogram had BFs well above 0.9 kHz, whereas for the other categories no relationship with BF was observed. Most units had a selectivity for ITD which was rather independent from frequency and absolute intensity level. In 62% of the units interaural time difference could be traded by interaural intensity difference. In most cases this so-called time-intensity trading could be explained by the intensity-latency characteristics of auditory nerve fibres. About 20% was sensitive to IID only and 5% to ITD only. A binaural model is proposed which is based on the intensity-rate and intensity-latency characteristics of auditory nerve fibres, the linear summation of excitatory and inhibitory post synaptic potentials in second order neurons, and spatiotemporal integration at the level of third order neurons. By variation of only a small number of parameters, namely strengths and time constants of the connectivities, the range of experimentally observed response patterns could be reproduced.

Effect on the partial least-squares prediction of yarn properties combining raman and infrared measurements and applying wavelength selection.

The combination of Raman and infrared spectroscopy on the one hand and wavelength selection on the other hand is used to improve the partial least-squares (PLS) prediction of seven selected yarn properties. These properties are important for on-line quality control during production. From 71 yarn samples, the Raman and infrared spectra are measured and reference methods are used to determine the selected properties. Making separate PLS models for all yarn properties using the Raman and infrared spectra, prior to wavelength selection, reveals that Raman spectroscopy outperforms infrared spectroscopy. If wavelength selection is applied, the PLS prediction error decreases and the correlation coefficient increases for all properties. However, a substantial wavelength selection effect is present for the infrared spectra compared to the Raman spectra. For the infrared spectra, wavelength selection results in PLS prediction errors comparable with the prediction performance of the Raman spectra prior to wavelength selection. Concatenating the Raman and infrared spectra does not enhance the PLS prediction performance, not even after wavelength selection. It is concluded that an infrared spectrometer, combined with a wavelength selection procedure, can be used if no (suitable) Raman instrument is available.

Visualisation and interpretation of Support Vector Regression models.

This paper introduces a technique to visualise the information content of the kernel matrix and a way to interpret the ingredients of the Support Vector Regression (SVR) model. Recently, the use of Support Vector Machines (SVM) for solving classification (SVC) and regression (SVR) problems has increased substantially in the field of chemistry and chemometrics. This is mainly due to its high generalisation performance and its ability to model non-linear relationships in a unique and global manner. Modeling of non-linear relationships will be enabled by applying a kernel function. The kernel function transforms the input data, usually non-linearly related to the associated output property, into a high dimensional feature space where the non-linear relationship can be represented in a linear form. Usually, SVMs are applied as a black box technique. Hence, the model cannot be interpreted like, e.g., Partial Least Squares (PLS). For example, the PLS scores and loadings make it possible to visualise and understand the driving force behind the optimal PLS machinery. In this study, we have investigated the possibilities to visualise and interpret the SVM model. Here, we exclusively have focused on Support Vector Regression to demonstrate these visualisation and interpretation techniques. Our observations show that we are now able to turn a SVR black box model into a transparent and interpretable regression modeling technique.

Detection and estimation of neural connectivity based on crosscorrelation analysis.

Crosscorrelation analysis of simultaneously recorded activity of pairs of neurons is a common tool to infer functional neural connectivity. The adequacy of crosscorrelation procedures to detect and estimate neural connectivity has been investigated by means of computer simulations of small networks composed of fairly realistic modelneurons. If the mean interval of neural firings is much larger than the duration of postsynaptic potentials, which will be the case in many central brain areas excitatory connections are easier to detect than inhibitory ones. On the other hand, inhibitory connections are revealed better if the mean firing interval is much smaller than post-synaptic potential duration. In general the effects of external stimuli and the effects of neural connectivity do not add linearly. Furthermore, neurons may exhibit a certain degree of timelock to the stimulus. For these reasons the commonly applied "shift predictor" procedure to separate stimulus and neural effects appears to be of limited value. In case of parallel direct and indirect neural pathways between two neurons crosscorrelation analysis does not estimate the direct connection but instead an effective connectivity, which reflects the combined influences of the parallel pathways.

Predicting nucleic acid torsion angle values using artificial neural networks.

By means of an error back-propagation artificial neural network, a new method to predict the torsion angles, chi, zeta and alpha from torsion angles delta, epsilon, beta and gamma for nucleic acid dinucleotides is introduced. To build a model, training sets and test sets of 163 and 81 dinucleotides, respectively, with known crystal structures, were assembled. With 7 hidden units in a three-layered network a model with good predictive ability is constructed. About 70 to 80% of the residuals for predicted torsion angles are smaller than 10 degrees. This means that such a model can be used to construct trial structures for conformational analysis that can be refined further. Moreover, when reasonable estimates for delta, epsilon, beta and gamma are extracted from COSY experiments, this procedure can easily be extended to predict torsion angles for structures in solution.

Parallel processing of chemical information in a local area network--II. A parallel cross-validation procedure for artificial neural networks.

This paper describes a parallel cross-validation (PCV) procedure, for testing the predictive ability of multi-layer feed-forward (MLF) neural networks models, trained by the generalized delta learning rule. The PCV program has been parallelized to operate in a local area computer network. Development and execution of the parallel application was aided by the HYDRA programming environment, which is extensively described in Part I of this paper. A brief theoretical introduction on MLF networks is given and the problems, associated with the validation of predictive abilities, will be discussed. Furthermore, this paper comprises a general outline of the PCV program. Finally, the parallel PCV application is used to validate the predictive ability of an MLF network modeling a chemical non-linear function approximation problem which is described extensively in the literature.

Parallel processing of chemical information in a local area network--III. Using genetic algorithms for conformational analysis of biomacromolecules.

Multi-dimensional nuclear magnetic resonance experiments are an excellent means of revealing the three-dimensional structure of biomacromolecules in solution. However, the search space in the conformational analysis of biomacromolecules, using multi-dimensional NMR data, is huge and complex. This calls for global optimization techniques with good sampling properties. This paper describes a genetic algorithm that optimizes the fit between (simulated) experimental two-dimensional Nuclear Overhauser Effect spectra and the corresponding calculated spectra for trial structures. This is a very computational intensive procedure. Speed-up of performance is achieved by parallelizing the algorithm, i.e. creating small subpopulations of trial structures, each of which can be processed on different processors. Good sampling behavior is obtained by initializing each subpopulation with its own random seed and the introduction of a migration operator. The latter replaces the best performing individual from one subpopulation with the worst performing individual from another subpopulation after a predetermined number of generations. A parallel genetic algorithm for the conformational analysis of nucleic acids is developed using the software package HYDRA. It is demonstrated that, for the data sets used in the study, a considerable reduction in computation time is obtained for the parallel genetic algorithm as compared to a sequential implementation, while the same optimal solutions are found.