Statistical Fusion Research Articles

Correction to "Formulating Spatially Varying Performance in the Statistical Fusion Framework" [Jun 12 1326-1336

In the above titled paper (ibid., vol. 31, no. 6, pp. 1326-1336, Jun. 2012), the first footnote should have included the following. This work was supported in part by NIH/NINDS 1R01EB006136, NIH/NINDS 1R01EB006193,NIH/NINDS 1R03EB012461, and NIH/NINDS 1R21NS064534.

A Novel Statistical Fusion Rule for Image Fusion and its Comparison in Non Subsampled Contourlet Transform Domain and Wavelet Domain

Image fusion produces a single fused image from a set of input images. A new method for image fusion is proposed based on Weighted Average Merging Method (WAMM) in the Non Subsampled Contourlet Transform (NSCT) domain. A performance analysis on various statistical fusion rules are also analysed both in NSCT and Wavelet domain. Analysis has been made on medical images, remote sensing images and multi focus images. Experimental results shows that the proposed method, WAMM obtained better results in NSCT domain than the wavelet domain as it preserves more edges and keeps the visual quality intact in the fused image.

Improved statistical image fusion method using a continuous-valued blur factor

Considering the reality of image formation as well as the statistical characteristics of contourlet coefficients, we propose a method for statistical image fusion in the contourlet domain. For high-frequency subbands, an image formation model was established that has a continuous-valued blur factor to reflect the actual imaging situation. According to this model, fused contourlet coefficients were estimated by use of an expectation-maximization (EM) algorithm. During the estimation, a contourlet hidden Markov tree model was adopted to grasp all dependencies among coefficients and aim for better estimation results. Because the blur factor is a continuous variable, we exploited an explicit expression to update the factor in the EM algorithm, which contributed to a decline in the number of iterations for convergence. Experimental results indicated that, especially for multifocus images, the proposed method provides more satisfying fusion results in terms of visual effects and objective evaluations than some typical fusion methods based on multiscale decomposition and some statistical fusion methods using a discrete blur factor.

Formulating Spatially Varying Performance in the Statistical Fusion Framework

To date, label fusion methods have primarily relied either on global [e.g., simultaneous truth and performance level estimation (STAPLE), globally weighted vote] or voxelwise (e.g., locally weighted vote) performance models. Optimality of the statistical fusion framework hinges upon the validity of the stochastic model of how a rater errs (i.e., the labeling process model). Hitherto, approaches have tended to focus on the extremes of potential models. Herein, we propose an extension to the STAPLE approach to seamlessly account for spatially varying performance by extending the performance level parameters to account for a smooth, voxelwise performance level field that is unique to each rater. This approach, Spatial STAPLE, provides significant improvements over state-of-the-art label fusion algorithms in both simulated and empirical data sets.

Foibles, follies, and fusion: Web-based collaboration for medical image labeling

Labels that identify specific anatomical and functional structures within medical images are essential to the characterization of the relationship between structure and function in many scientific and clinical studies. Automated methods that allow for high throughput have not yet been developed for all anatomical targets or validated for exceptional anatomies, and manual labeling remains the gold standard in many cases. However, manual placement of labels within a large image volume such as that obtained using magnetic resonance imaging (MRI) is exceptionally challenging, resource intensive, and fraught with intra- and inter-rater variability. The use of statistical methods to combine labels produced by multiple raters has grown significantly in popularity, in part, because it is thought that by estimating and accounting for rater reliability estimates of the true labels will be more accurate. This paper demonstrates the performance of a class of these statistical label combination methodologies using real-world data contributed by minimally trained human raters. The consistency of the statistical estimates, the accuracy compared to the individual observations, and the variability of both the estimates and the individual observations with respect to the number of labels are presented. It is demonstrated that statistical fusion successfully combines label information using data from online (Internet-based) collaborations among minimally trained raters. This first successful demonstration of a statistically based approach using minimally trained raters opens numerous possibilities for very large scale efforts in collaboration. Extension and generalization of these technologies for new applications will certainly present fascinating areas for continuing research.

Robust statistical label fusion through COnsensus Level, Labeler Accuracy, and Truth Estimation (COLLATE).

Segmentation and delineation of structures of interest in medical images is paramount to quantifying and characterizing structural, morphological, and functional correlations with clinically relevant conditions. The established gold standard for performing segmentation has been manual voxel-by-voxel labeling by a neuroanatomist expert. This process can be extremely time consuming, resource intensive and fraught with high inter-observer variability. Hence, studies involving characterizations of novel structures or appearances have been limited in scope (numbers of subjects), scale (extent of regions assessed), and statistical power. Statistical methods to fuse data sets from several different sources (e.g., multiple human observers) have been proposed to simultaneously estimate both rater performance and the ground truth labels. However, with empirical datasets, statistical fusion has been observed to result in visually inconsistent findings. So, despite the ease and elegance of a statistical approach, single observers and/or direct voting are often used in practice. Hence, rater performance is not systematically quantified and exploited during label estimation. To date, statistical fusion methods have relied on characterizations of rater performance that do not intrinsically include spatially varying models of rater performance. Herein, we present a novel, robust statistical label fusion algorithm to estimate and account for spatially varying performance. This algorithm, COnsensus Level, Labeler Accuracy and Truth Estimation (COLLATE), is based on the simple idea that some regions of an image are difficult to label (e.g., confusion regions: boundaries or low contrast areas) while other regions are intrinsically obvious (e.g., consensus regions: centers of large regions or high contrast edges). Unlike its predecessors, COLLATE estimates the consensus level of each voxel and estimates differing models of observer behavior in each region. We show that COLLATE provides significant improvement in label accuracy and rater assessment over previous fusion methods in both simulated and empirical datasets.

A Novel and Simple Statistical Fusion Method for User Authentication through Keystroke Features

This paper utilizes a statistical fusion method to extract the characteristic information for keystrokebased authentication (KA) systems to verify users’ identities. The keystroke data is based on the time instances of pressing and releasing a key, and five features based on the time periods are calculated using these data. In the experiment, nineteen users participated as the legitimate users and each account was attacked by between 62 and 82 impostors. The average false acceptance rate is 1.035% and the false rejection rate is 0%. These rates are both competitive with other researches. The proposed method can be used as the classifier in password-based authentication systems to enhance their security.

Generalized Evidential Processing in Multiple Simultaneous Threat Detection in UNIX

In this paper, a hybrid identity fusion model at decision level is proposed for Simultaneous Threat Detection Systems. The hybrid model is comprised of mathematical and statistical data fusion engines; Dempster Shafer, Extended Dempster and Generalized Evidential Processing (GEP). Simultaneous Threat Detection Systems improve threat detection rate by 39%. In terms of efficiency and performance, the comparison of 3 inference engines of the Simultaneous Threat Detection Systems showed that GEP is the better data fusion model. GEP increased precision of threat detection from 56% to 95%. Furthermore, set cover packing was used as a middle tier data fusion tool to discover the reduced size groups of threat data. Set cover provided significant improvement and reduced threat population from 2272 to 295, which helped in minimizing the processing complexity of evidential processing cost and time in determining the combined probability mass of proposed Multiple Simultaneous Threat Detection System. This technique is particularly relevant to on-line and Internet dependent applications including portals.

Statistical fusion of lidar, InSAR, and optical remote sensing data for forest stand height characterization: A regional‐scale method based on LVIS, SRTM, Landsat ETM+, and ancillary data sets

A method is presented to characterize forest stand heights in a 110,000 km2 region in the eastern United States surrounding the Chesapeake Bay area, driven by a statistical fusion model solely based on remote sensing data. The predicted map was tested against ground survey data from the Forest Inventory and Analysis (FIA) plot network. Input data to the model were 2003 medium footprint lidar data from the Laser Vegetation Imaging Sensor (LVIS) sensor, interferometric radar data from the 2000 Shuttle Radar Topography Mission (SRTM), 1999–2001 Landsat ETM+ data, and ancillary data sets of land cover and canopy density developed for the 2001 National Land Cover Database. In the presented approach, the interferometric synthetic aperture radar (InSAR), optical, and ancillary data sets were masked to the forested areas of the study region and used to segment the raster data stack. The generated image objects closely represented quasi‐homogenous forest stands. For a small region in the study area covered by an LVIS acquisition, LVIS lidar data were then used within the established segments to extract lidar‐based mean forest stand heights. Subsequently these LVIS mean stand heights were used as the response variable to the statistical prediction model (randomForest) which had segment‐based metrics like mean InSAR height (derived from SRTM minus ground digital elevation model data from the National Elevation Data set), mean optical reflectance (derived from Landsat ETM+ Tassled Cap Data), and ancillary metrics as predictive variables. The model developed over the area where LVIS data were available was then applied to map the entire study region. Independent validation of the model was performed in two ways. First, splitting of the model data stack into training and independent testing populations, i.e., testing on LVIS data. This test was deemed to describe the model performance within the LVIS swath. Second, predicted heights were compared to plot height metrics derived from FIA data in the entire study region, thus testing the validity of the model across the larger study area. Results, which are somewhat tampered by the time disconnect between the various data collections, showed the validity and usefulness of this approach. Independent LVIS testing resulted in a correlation coefficient r = 0.83 with an RMSE of 3.0 m (9% error), independent FIA data tested with r = 0.71 with an RMSE of 4.4 m (13% error).

Statistical fusion of two-scale images of porous media

The reconstruction of the architecture of void space in porous media is a challenging task, since porous media contain pore structures at multiple scales. Whereas past methods have been limited to producing samples with matching statistical behavior, the patterns of grey-level values in a measured sample actually say something about the unresolved details, thus we propose a statistical fusion framework for reconstructing high-resolution porous media images from low-resolution measurements. The proposed framework is based on a posterior sampling approach in which information obtained by low-resolution (MRI or X-ray) measurements is combined with prior models inferred from high-resolution microscopic data, typically 2D. In this paper, we focus on two-scale reconstruction tasks in which the measurements resolve only the large scale structures, leaving the small-scale to be inferred. The evaluation of the results generated by the proposed method shows the strong ability of the proposed method in reconstructing fine-scale structures positively correlated with the underlying ground truth. Comparing our method with the recent method of Okabe and Blunt [12], in which the measurements are also used in the reconstruction, we conclude that our method is more robust to the resolution of the measurement, and more closely matches the underlying fine-scale field.

Time-Scale Feature Extractions for Emotional Speech Characterization

Emotional speech characterization is an important issue for the understanding of interaction. This article discusses the time-scale analysis problem in feature extraction for emotional speech processing. We describe a computational framework for combining segmental and supra-segmental features for emotional speech detection. The statistical fusion is based on the estimation of local a posteriori class probabilities and the overall decision employs weighting factors directly related to the duration of the individual speech segments. This strategy is applied to a real-world application: detection of Italian motherese in authentic and longitudinal parent–infant interaction at home. The results suggest that short- and long-term information, respectively, represented by the short-term spectrum and the prosody parameters (fundamental frequency and energy) provide a robust and efficient time-scale analysis. A similar fusion methodology is also investigated by the use of a phonetic-specific characterization process. This strategy is motivated by the fact that there are variations across emotional states at the phoneme level. A time-scale based on both vowels and consonants is proposed and it provides a relevant and discriminant feature space for acted emotion recognition. The experimental results on two different databases Berlin (German) and Aholab (Basque) show that the best performance are obtained by our phoneme-dependent approach. These findings demonstrate the relevance of taking into account phoneme dependency (vowels/consonants) for emotional speech characterization.

Dynamically optimized spatiotemporal prioritization for spectrum sensing in cooperative cognitive radio

In this paper, an enhanced cooperative, statistics-driven spectrum sensing algorithm, called Dynamically Optimized Spatiotemporal Prioritization (DOSP), is developed for improving spectrum sensing efficiency in the media access control (MAC) layer of cognitive radio (CR) systems. The target of the DOSP algorithm is to improve spectrum sensing efficiency and achieve better spectrum access opportunities by prioritizing channels for fine sensing. The sensing priority is determined dynamically and intelligently based on an optimal statistical fusion that jointly considers both the local statistics obtained by the individual cognitive radios as well as the long-term spatiotemporal statistics obtained by other cognitive radios in the network. As such, the individual cognitive radio peers work together to get the most out of available spectrum opportunities. Numerical results demonstrate that the proposed DOSP algorithm is capable of achieving better performance compared with recently reported cooperative spectrum sensing methods in terms of overhead and percentage of missed spectrum opportunities. Furthermore, results show that the DOSP algorithm is more robust to the environment of low cognitive radio densities than that by using other state-of-the-art cooperative spectrum sensing methods.

A statistical image fusion algorithm for NDE applications

A statistical image fusion algorithm for NDE applications

Grid Enabling Empirical Economics: A Microdata Application

This article discusses the use of Grid technology to integrate the data, computation and presentation elements of an empirical economic modelling process. We achieve this by using a form of statistical data fusion developed in the poverty mapping literature to address a substantive issue: determining United Kingdom ethnic minority welfare. Elements of this methodology appear to be well suited to such grid-enablement, and we present and illustrate our implementation using the context of this microdata application.

A regression tree-based method for integrating land-cover and land-use data collected at multiple scales

As data sets of multiple types and scales proliferate, it will be increasingly important to be able to flexibly combine them in ways that retain relevant information. A case in point is Amazonia, a large, data-poor region where most whole-basin data sets are limited to understanding land cover interpreted through a variety of remote sensing techniques and sensors. A growing body of work, however, indicates that the future state of much of Amazonia depends on the land use to which converted areas are put, but land use in the tropics is difficult to assess from remotely sensed data alone. An earlier paper developed new snapshots of agricultural land use in this region using a statistical fusion of satellite data and agricultural census data, an underutilized ancillary data source available across Amazonia. The creation of these land-use maps, which have the spatial detail of a satellite image and the attribute information of an agricultural census, required the development of a new statistical technique for merging data sets at different scales and of fundamentally different data types. Here we describe and assess this nonlinear technique, which reinterprets existing land cover classifications by determining what categories are most highly related to the polygon land-use data across the study area. Although developed for this region, the technique appears to hold broad promise for the systematic fusion of multiple data sets that are closely related but of different origins.

Fusion of multisensor multitemporal satellite data for land cover mapping

Efficient integration of remote sensing information with different temporal, spectral and spatial resolutions is important for accurate land cover mapping. A new temporal fusion classification (TFC) model is presented for land cover classification, based on statistical fusion of multitemporal satellite images. In the proposed model, the temporal dependence of multitemporal images is taken into account by estimating transition probabilities from the change pattern of a vegetation dynamics indicator (VDI). Extension of this model is applicable to Synthetic Aperture Radar (SAR) images and integration of multisensor multitemporal satellite images, concerning both temporal attributes and reliability of multiple data sources. The feasibility of the new method is verified using multitemporal Landsat Thematic Mapper (TM) and ERS SAR satellite images, and experimental results show improved performance over conventional methods.

A method for statistical data fusion on a per‐field basis

The aim of this study is to develop a new method for data fusion, in a statistical manner, on a per‐field basis that can be applied before a per‐field classification for increasing per‐field classification accuracy. The method developed for statistical data fusion is a model‐based data fusion method which utilizes nested mixture distribution modelling and integrates two multiresolution and multispectral image data each from different wavelengths of different spectral bands of different imaging sensors with different resolutions and having a different number of spectral bands. The statistical data fusion process involves two successive steps.

Fusion of reduced-rank TMA estimates

The problem of bearings-only target motion analysis (TMA) in incompletely observable situations is considered. Both target and observer trajectories are rectilinear uniform motions (RUMs). A specific method is first established for the estimation of such reduced-rank TMA (RR-TMA). Then, the statistical fusion of two (or more) RR-TMA estimates into a complete observable TMA is established. Remarkably, this second stage is analytical. Finally the fusion algorithm is shown to be efficient for a reasonable domain of measurement errors. Applications are listed among radar and sonar tracking problems. They are specially encountered in large-scale multiplatform situations where combinatorial testing of multiple hypotheses is a big issue, or in long-term integration of moving targets where the dimension reduction of the search-grid is critical.

The competition between fusion-fission and deeply inelastic reactions in the medium mass systems

The competition between fusion-fission and deeply inelastic reactions in the Cl + Ni system has been studied by investigating the exit channelQ-value dependence of theZ-distributions of fission-like fragments. TheZ-, kinetic energy- and angular distributions of the fission-like fragments produced in the37Cl +64Ni system atElab=170, 186 MeV and in the28Si +74Ge system atElab=176 MeV have been measured by counter telescopes. TheZ-distribution of the37Cl +64Ni system was found to be essentially the same as that of the35Cl +62Ni system at the same bombarding energy. It is shown that this result can not be explained by the statistical fusion fission model but is consistent with a deeply inelastic model. A systematic study of the fission-like phenomenon in this mass region indicates that the maximum angular momenta for fusion reactions as well as the minimum angular momenta for DIC can be reproduced by the Bass model in the sliding limit (f=1.0) while the maximum angular momenta for fully energy-damped deeply inelastic reactions are governed by the strong interaction radius as predicted by the successive critical distance fusion model. The significance of these results are discussed.