Reduction of multi-dimensional laboratory data to a two-dimensional plot: a novel technique for the identification of laboratory error.

BACKGROUND: The clinical laboratory generates large amounts of patient-specific data. Detection of errors that arise during pre-analytical, analytical, and post-analytical processes is difficult. We performed a pilot study, utilizing a multidimensional data reduction technique, to assess the utility of this method for identifying errors in laboratory data. METHODS: We evaluated 13,670 individual patient records collected over a 2-month period from hospital inpatients and outpatients. We utilized those patient records that contained a complete set of 14 different biochemical analytes. We used two-dimensional generative topographic mapping to project the 14-dimensional record to a two-dimensional space. RESULTS AND CONCLUSIONS: The use of a two-dimensional generative topographic mapping technique to plot multi-analyte patient data as a two-dimensional graph allows for the rapid identification of potentially anomalous data. Although we performed a retrospective analysis, this technique has the benefit of being able to assess laboratory-generated data in real time, allowing for the rapid identification and correction of anomalous data before they are released to the physician. In addition, serial laboratory multi-analyte data for an individual patient can also be plotted as a two-dimensional plot. This tool might also be useful for assessing patient wellbeing and prognosis.

Gaussian mean-shift is an EM algorithm.

The mean-shift algorithm, based on ideas proposed by Fukunaga and Hostetler [16], is a hill-climbing algorithm on the density defined by a finite mixture or a kernel density estimate. Mean-shift can be used as a nonparametric clustering method and has attracted recent attention in computer vision applications such as image segmentation or tracking. We show that, when the kernel is Gaussian, mean-shift is an expectation-maximization (EM) algorithm and, when the kernel is non-Gaussian, mean-shift is a generalized EM algorithm. This implies that mean-shift converges from almost any starting point and that, in general, its convergence is of linear order. For Gaussian mean-shift, we show: 1) the rate of linear convergence approaches 0 (superlinear convergence) for very narrow or very wide kernels, but is often close to 1 (thus, extremely slow) for intermediate widths and exactly 1 (sublinear convergence) for widths at which modes merge, 2) the iterates approach the mode along the local principal component of the data points from the inside of the convex hull of the data points, and 3) the convergence domains are nonconvex and can be disconnected and show fractal behavior. We suggest ways of accelerating mean-shift based on the EM interpretation.

A computational model for the development of multiple maps in primary visual cortex.

Primary visual cortex contains multiple maps of features of the visual scene, including visual field position, orientation, direction, ocular dominance and spatial frequency. The complex relationships between these maps provide clues to the strategies the cortex uses for representing and processing information. Here we simulate the combined development of all these map systems using a computational model, the elastic net. We show that this model robustly produces combined maps of these four variables that bear a close resemblance to experimental maps. In addition we show that the experimentally observed effects of monocular deprivation and single-orientation rearing can be reproduced in this model, and we make some testable predictions. These results provide strong support for the hypothesis that cortical representations attempt to optimize a trade-off between coverage and continuity.

Influence of lateral connections on the structure of cortical maps.

Maps of ocular dominance and orientation in primary visual cortex have a highly characteristic structure. The factors that determine this structure are still largely unknown. In particular, it is unclear how short-range excitatory and inhibitory connections between nearby neurons influence structure both within and between maps. Using a generalized version of a well-known computational model of visual cortical map development, we show that the number of excitatory and inhibitory oscillations in this interaction function critically influences map structure. Specifically, we demonstrate that functions that oscillate more than once do not produce maps closely resembling those seen biologically. This strongly suggests that local lateral connections in visual cortex oscillate only once and have the form of a Mexican hat.

Are visual cortex maps optimized for coverage?

The elegant regularity of maps of variables such as ocular dominance, orientation, and spatial frequency in primary visual cortex has prompted many people to suggest that their structure could be explained by an optimization principle. Up to now, the standard way to test this hypothesis has been to generate artificial maps by optimizing a hypothesized objective function and then to compare these artificial maps with real maps using a variety of quantitative criteria. If the artificial maps are similar to the real maps, this provides some evidence that the real cortex may be optimizing a similar function to the one hypothesized. Recently, a more direct method has been proposed for testing whether real maps represent local optima of an objective function (Swindale, Shoham, Grinvald, Bonhoeffer, & Hübener, 2000). In this approach, the value of the hypothesized function is calculated for a real map, and then the real map is perturbed in certain ways and the function recalculated. If each of these perturbations leads to a worsening of the function, it is tempting to conclude that the real map is quite likely to represent a local optimum of that function. In this article, we argue that such perturbation results provide only weak evidence in favor of the optimization hypothesis.