Biomedical subjects

Paul A Watters

Publications and source records attributed to Paul A Watters.

5 recordsLinked to original sources

PubMed2006

Are sparse-coding simple cell receptive field models physiologically plausible?

Olshausen and Field (1996) developed a simple cell receptive field model for natural scene processing in V1, based on unsupervised learning and non-orthogonal basis function optimization of an overcomplete representation of visual space. The model was originally tested with an ensemble of whitened natural scenes, simulating pre-cortical filtering in the retinal ganglia and lateral geniculate nucleus, and the basis functions qualitatively resembled the orientation-specific responses of V1 simple cells in the spatial domain. In this study, the quantitative tuning responses of the basis functions in the spectral domain are estimated using a Gaussian model, to determine their goodness-of-fit to the known bandwidths of simple cells in primate V1. Five simulation experiments which examined key features of the model are reported: changing the size of the basis functions; using a complete versus over-complete representation; changing the sparseness factor; using a variable learning rate; and mapping the basis functions with a whitening spatial function. The key finding of this study is that across all image themes, basis function sizes, number of basis functions, sparseness factors and learning rates, the spatial-frequency tuning did not closely resemble that of primate area 17 -- the model results more closely resembled the unclassified cat neurones of area 19 with a single exception, and not area 17 as predicted.

Action Potentials↗

PubMed2005

Selecting parameters for phase space reconstruction of the electrocorticogram (ECoG).

The selection of parameters for phase space reconstruction of empirically observed data has been a source of criticism when estimating the correlation dimension (D2) from observed data rather than from the solution of differential equations, when analyzing noisy and potentially non-stationary signals, such as the electroencephalogram (EEG). The largely arbitrary selection of the time-delay reconstruction (T) of temporal dynamics, and for the embedding (M) of these series, has been widely criticized. This study adopted an analytic and statistical framework within which the scaling behavior of D2 with respect to T and M, could be examined over five data lengths (N = 4096, 8192, 12288, 16384, and 20480) over an 8 x 8 grid of cat EEG. It was found that D2 was invariant over all data lengths only within a very narrow T range (T = 10-16) for M = 4. A statistically significant T by M interaction was found using multiple analysis of variance, with D2 being highly correlated over T as a function of M. Finally, an examination of phase-randomized surrogates indicated that statistically significant differences existed between EEG and phase-randomized surrogates over all data lengths, with time delays (T = 10-16), indicating that the D2 for EEG is phase-dependent when it is invariant with respect to data length. The implications of these findings are discussed with respect to current models of ECoG generation, and their implication with respect to the integration in the brain.

Analysis of Variance↗

PubMed2004

A method for estimating long-range power law correlations from the electroencephalogram.

Recent research has found long-range electroencephalogram (EEG) power law correlations, indicating time scale invariance. However, the EEG is also rather noisy, displaying short-term decorrelation like white noise--i.e., what is scale invariant at one time period may disappear in the next. The paradoxical combination of short-range divergence, but long-range correlations, suggests that any long-range correlations detected in one sample may be spurious, since they could be related to amplitude fluctuations. To overcome this problem, this paper suggests a new technique for analysing EEG signals segmented by zero-crossings, using detrended fluctuation analysis (DFA), evaluated across two time periods (TIME) and different sites (SITE). A mean scaling exponent across all subjects and sites of alpha = 0.67 was observed. MANOVA analysis indicates no significant main effect for TIME or interaction with SITE, suggesting that the zero-crossing method may be successful in determining the fractal nature of EEG dynamics across relatively long time scales.

Cerebral Cortex↗

PubMed2003

Estimating distributed coding efficiency in orthogonal models of facial processing.

Orthogonal facial processing models attempt to mimic the local decomposition performed in the visual cortex by simple cell receptive fields. The purpose of this study was to investigate how the neurophysiological validity of orthogonal models of facial processing could be improved by implementing a "whitening" filter, based on current knowledge of similar filtering that occurs in the retina. By using a metric known as the "distributed coding efficiency index" (DCE), this study demonstrates that an orthogonal facial processing model significantly increased coding efficiency when a low-pass, zero-phase whitening filter was applied. The extent to which orthogonal decomposition of filtered data represents a realistic V1 model is discussed.

Face↗

PubMed2002

Competition, inhibition, and semantic judgment errors in Parkinson's disease.

Semantic processing errors are symptoms of an up-regulation (schizophrenia) or degradation (Parkinsonism) of dopaminergic pathways. A recent connectionist model attributed errors in the schizophrenic processing of context to increased gain in competitive neural processes. This study extends this "gain hypothesis" by comparing the sensitivity to reduced gain of a simulation of semantic route activation to characteristic semantic judgment errors made by Parkinson's patients in an open search task. Under normal gain conditions, the dominant sense of polysemous words "wins" through competition and lateral inhibition at the word sense level (beta(inh)). For words with very different sense frequencies, decreasing gain by increasing beta(inh) resulted in the dominant word sense winning; however, for words with similar sense frequencies, increasing beta(inh) resulted in the dominant word sense winning only for low to moderate values. At high levels, no clear winner emerged after 200 epochs, with the least dominant sense reaching the maximum activation value. These results are discussed in the context of the Yerkes-Dodson Law, which may provide a theoretical basis for understanding normal and impaired semantic performance in catecholaminergic disorders.

Aphasia↗