Spatial quantitative analysis of fluorescently labeled nuclear structures: problems, methods, pitfalls.

The vast majority of microscopic data in biology of the cell nucleus is currently collected using fluorescence microscopy, and most of these data are subsequently subjected to quantitative analysis. The analysis process unites a number of steps, from image acquisition to statistics, and at each of these steps decisions must be made that may crucially affect the conclusions of the whole study. This often presents a really serious problem because the researcher is typically a biologist, while the decisions to be taken require expertise in the fields of physics, computer image analysis, and statistics. The researcher has to choose between multiple options for data collection, numerous programs for preprocessing and processing of images, and a number of statistical approaches. Written for biologists, this article discusses some of the typical problems and errors that should be avoided. The article was prepared by a team uniting expertise in biology, microscopy, image analysis, and statistics. It considers the options a researcher has at the stages of data acquisition (choice of the microscope and acquisition settings), preprocessing (filtering, intensity normalization, deconvolution), image processing (radial distribution, clustering, co-localization, shape and orientation of objects), and statistical analysis.

Assessing DTI data quality using bootstrap analysis.

Diffusion tensor imaging (DTI) is an established method for characterizing and quantifying ultrastructural brain tissue properties. However, DTI-derived variables are affected by various sources of signal uncertainty. The goal of this study was to establish an objective quality measure for DTI based on the nonparametric bootstrap methodology. The confidence intervals (CIs) of white matter (WM) fractional anisotropy (FA) and Clinear were determined by bootstrap analysis and submitted to histogram analysis. The effects of artificial noising and edge-preserving smoothing, as well as enhanced and reduced motion were studied in healthy volunteers. Gender and age effects on data quality as potential confounds in group comparison studies were analyzed. Additional noising showed a detrimental effect on the mean, peak position, and height of the respective CIs at 10% of the original background noise. Inverse changes reflected data improvement induced by edge-preserving smoothing. Motion-dependent impairment was also well depicted by bootstrap-derived parameters. Moreover, there was a significant gender effect, with females displaying less dispersion (attributable to elevated SNR). In conclusion, the bootstrap procedure is a useful tool for assessing DTI data quality. It is sensitive to both noise and motion effects, and may help to exclude confounding effects in group comparisons.

Fiber tracking from DTI using linear state space models: detectability of the pyramidal tract.

Diffusion tensor imaging (DTI) is an emerging and promising tool to provide information about the course of white matter fiber tracts in the human brain. Based on specific acquisition schemes, diffusion tensor data resemble local fiber orientations allowing for a reconstruction of the fiber bundles. Current techniques to calculate fascicles range from simple heuristic tracking solutions to Bayesian and differential equations approaches. Most methods are based only on local diffusion information, often resulting in bending or kinking fiber paths in voxels with reduced diffusion properties. In this article we present a new tracking approach based on linear state space models encompassing an inherent smoothness criterion to avoid too wiggly tracked fiber bundles. The new technique will be described formally and tested on simulated and real data. The performance tests are focused on the pyramidal tract, where we employed a test-retest study and a group comparison in healthy subjects. Anatomical course was confirmed in a patient with selective degeneration of the pyramidal tract. The potential of the presented technique for improved neurosurgical planning is demonstrated by visualization of a tumor-induced displacement of the motor pathways. The paper closes with a thorough discussion of perspectives and limitations of the new tracking approach.

Bayesian modeling of the hemodynamic response function in BOLD fMRI.

In functional magnetic resonance imaging (fMRI), modeling the complex link between neuronal activity and its hemodynamic response via the neurovascular coupling requires an elaborate and sensitive response model. Methods based on physiologic assumptions as well as direct, descriptive models have been proposed. The focus of this study is placed on such a direct approach that allows for a robust pixelwise determination of hemodynamic characteristics, such as time to peak or the poststimulus undershoot. A Bayesian procedure is presented that can easily be adapted to different hemodynamic properties in question and can be estimated without numerical problems known from nonlinear optimization algorithms. The usefulness of the model is demonstrated by thorough analyzes of the poststimulus undershoot in visual and acoustic stimulation paradigms. Further, we show the capability of this approach to improve analysis of fMRI data in altered hemodynamic conditions.

Bayesian spatiotemporal inference in functional magnetic resonance imaging.

Mapping of the human brain by means of functional magnetic resonance imaging (fMRI) is an emerging field in cognitive and clinical neuroscience. Current techniques to detect activated areas of the brain mostly proceed in two steps. First, conventional methods of correlation, regression, and time series analysis are used to assess activation by a separate, pixelwise comparison of the fMRI signal time courses to the reference function of a presented stimulus. Spatial aspects caused by correlations between neighboring pixels are considered in a separate second step, if at all. The aim of this article is to present hierarchical Bayesian approaches that allow one to simultaneously incorporate temporal and spatial dependencies between pixels directly in the model formulation. For reasons of computational feasibility, models have to be comparatively parsimonious, without oversimplifying. We introduce parametric and semiparametric spatial and spatiotemporal models that proved appropriate and illustrate their performance applied to visual fMRI data.

Dynamic models in fMRI.

Most statistical methods for assessing activated voxels in fMRI experiments are based on correlation or regression analysis. In this context, the main assumptions are that the baseline can be described by a few known basis functions or variables and that the effect of the stimulus, i.e., the activation, stays constant over time. As these assumptions are in many cases neither necessary nor correct, a new dynamic approach that does not depend on those assumptions will be presented. This allows for simultaneous nonparametric estimation of the baseline and, as an important feature, of time-varying effects of stimulation. This method of estimating the stimulus related areas of the brain furthermore provides the possibility to analyze the temporal and spatial evolution of the activation within an fMRI experiment.

Modelling and exploring human sleep with event history analysis.

In this paper we propose the use of statistical models of event history analysis for investigating human sleep. These models provide appropriate tools for statistical evaluation when sleep data are recorded continuously over time or on a fine time grid, and are classified into sleep stages such as REM and nonREM as defined by Rechtschaffen and Kales (1968). In contrast to conventional statistical procedures, event history analysis makes full use of the information contained in sleep data, and can therefore provide new insights into non-stationary properties of sleep. Probabilities of or intensities for transitions between sleep stages are the basic quantities for characterising sleep processes. The statistical methods of event history analysis aim at modelling and estimating these intensities as functions of time, taking into account individual sleep history and assessing the influence of factors of interest, such as hormonal secretion. In this study we suggest the use of non-parametric approaches to reveal unknown functional forms of transition intensities and to explore time-varying and non-stationary effects. We then apply these techniques in a study of 30 healthy male volunteers to assess the mean population intensity and the effects of plasma cortisol concentration on the transition between selected sleep stages as well as the influence of elapsed time in a current REM period on the intensity for a transition to nonREM. The most interesting findings are that (a) the intensity of the nonREM-to-REM transitions after sleep onset in young men shows a periodicity which is similar to that of nonREM/REM cycles; (b) 30-45 min after sleep onset, young men reveal a great propensity to pass from light sleep (stages 1 or 2) into slow-wave sleep (SWS) (stages 3 or 4); (c) high cortisol levels imposed additional impulses on the transition intensity of (i) wake to sleep around 2 h after sleep onset, (ii) nonREM to REM around 6 h later, (iii) stage 1 or stage 2 sleep to SWS around 2, 4 and 6 h later and (iv) SWS to stage 1 or stage 2 sleep about 2 h later. Moreover, high cortisol concentrations at the beginning of REM periods favoured the change to nonREM sleep, whereas later their influence on a nonREM change became weak and weaker. As sleep data are also available as event-oriented data in many studies in sleep research, event history analysis applied additionally to conventional statistical procedures, such as regression analysis or analysis of variance, could help to acquire more information and knowledge about the mechanisms behind the sleep process.

An application of isotonic longitudinal marginal regression to monitoring the healing process.

This paper discusses marginal regression for repeated ordinal measurements that are isotonic over time. Such data are often observed in longitudinal studies on healing processes in which, due to recovery, the status of patients only improves or remains the same. We show how this prior information can be used to construct appropriate and parsimoniously parametrized marginal models. As a second aspect, we also incorporate nonparametric fitting of covariate effects via a penalized quasi-likelihood or general estimating equation approach. We illustrate our methods by an application to sports-related injuries.