Oscar Yanez-Suarez

Data-driven brain MRI segmentation supported on edge confidence and a priori tissue information

Brain magnetic resonance imaging segmentation is accomplished in this work by applying nonparametric density estimation, using the mean shift algorithm in the joint spatial-range domain. The quality of the class boundaries is improved by including an edge confidence map, that represents the confidence of truly being in the presence of a border between adjacent regions; an adjacency graph is then constructed with the labeled regions, and analyzed and pruned to merge adjacent regions. In order to assign image regions to a cerebral tissue type, a spatial normalization between image data and standard probability maps is carried out, so that for each structure a maximum a posteriori probability criterion is applied. The method was applied to synthetic and real images, keeping all parameters constant throughout the process for each type of data. The combination of region segmentation and edge detection proved to be a robust technique, as adequate clusters were automatically identified, regardless of the noise level and bias. In a comparison with reference segmentations, average Tanimoto indexes of 0.90-0.99 were obtained for synthetic data and of 0.59-0.99 for real data, considering gray matter, white matter, and background.

Coupling of radial-basis network and active contour model for multispectral brain MRI segmentation

Magnetic resonance (MR) has been accepted as the reference image study in the clinical environment. The development of new sequences has allowed obtaining diverse images with high clinical importance and whose interpretation requires their joint analysis (multispectral MRI). Recent approaches to segment MRI point toward the definition of hybrid models, where the advantages of region and contour-based methods can be exploited to look for the integration or fusion of information, thus enhancing the performance of the individual approaches. Following this perspective, a hybrid model for multispectral brain MRI segmentation is presented. The model couples a segmenter, based on a radial basis network (RBFNNcc), and an active contour model, based on a cubic spline active contour (CSAC) interpolation. Both static and dynamic coupling of RBFNNcc and CSAC are proposed; the RBFNNcc stage provides an initial contour to the CSAC; the initial contour is optimally sampled with respect to its curvature variations; multispectral information and a restriction term are included into the CSAC energy equation. Segmentations were compared to a reference stack, indicating high-quality performance as measured by Tanimoto indexes of 0.74/spl plusmn/0.08.

Baroreflex sensitivity assessment and heart rate variability: relation to maneuver and technique

In the present study, we examined two baroreflex sensitivity (BRS) issues that remain uncertain: the differences among diverse BRS assessment techniques and the association between BRS and vagal outflow. Accordingly, the electrocardiogram and non-invasive arterial pressure were recorded in 27 healthy subjects, during supine with and without controlled breathing, standing, exercise, and recovery conditions. Vagal outflow was estimated by heart rate variability indexes, whereas BRS was computed by alpha-coefficient, transfer function, complex demodulation in low- and high-frequency bands, and by sequence technique. Our results indicated that only supine maneuvers showed significantly greater BRS values over the high frequency than in the low-frequency band. For maneuvers at the same frequency region, supine conditions presented a larger number of significant differences among techniques. The plots between BRS and vagal measures depicted a funnel-shaped relationship with significant log–log correlations (r=0.880–0.958). Very short latencies between systolic pressure and RR interval series in high-frequency band and strong log–log correlations between frequency bands were found. Higher variability among different baroreflex measurements was associated with higher level of vagal outflow. Methodological assumptions for each technique seem affected by non-baroreflex variation sources, and a modified responsiveness of vagal motoneurons due to distinct stimulation levels for each maneuver was suggested. Thus, highest vagal outflows corresponded to greatest BRS values, with maximum respiratory effect for the high-frequency band values. In conclusion, BRS values and differences across the tested techniques were strongly related to the vagal outflow induced by the maneuvers.

An FPGA Implementation of Linear Kernel Support Vector Machines

This paper describes preliminary performance results of a reconfigurable hardware implementation of a support vector machine classifier, aimed at brain-computer interface applications, which require real-time decision making in a portable device. The main constraint of the design was that it could perform a classification decision within the time span of an evoked potential recording epoch of 300 ms, which was readily achieved for moderate-sized support vector sets. Regardless of its fixed-point implementation, the FPGA-based model achieves equivalent classification accuracies to those of its software-based, floating-point counterparts

Robust Nonparametric Segmentation of Infarct Lesion from Diffusion-Weighted MR Images

Magnetic Resonance Imaging (MRI) is increasingly used for the diagnosis and monitoring of neurological disorders. In particular Diffusion-Weighted MRI (DWI) is highly sensitive in detecting early cerebral ischemic changes in acute stroke. Cerebral infarction lesion segmentation from DWI is accomplished in this work by applying nonparametric density estimation. The quality of the class boundaries is improved by including an edge confidence map, that is the confidence of truly being in the presence of a border between adjacent regions. The adjacency graph, that is constructed with the label regions, is analyzed and pruned to merge adjacent regions. The method was applied to real images, keeping all parameters constant throughout the process for each data set. The combination of region segmentation and edge detection proved to be a robust automatic technique of segmentation from DWI images of cerebral infarction regions in acute ischemic stroke. In a comparison with the reference infarct lesions segmentation, the automatic segmentation presented a significant correlation (r = 0.935), and an average Tanimoto index of 0.538.

Single Trial P300 detection based on the Empirical Mode Decomposition

We present a new method for single trial detection of P300 evoked responses. The features used to classify are the coefficients of a least-squares fit of a single EEG epoch to the intrinsical mode functions of an empirical mode decomposition of the averaged event response from a P300 training set. Support vector machines with a linear kernel are used to classify the epochs and receiver operating characteristic analysis is used to evaluate our methods performance

Segmenting Brain MRI using Adaptive Mean Shift

To delineate arbitrarily shaped clusters in a complex multimodal feature space, such as the brain MRI intensity space, often requires kernel estimation techniques with locally adaptive bandwidths, such as the adaptive mean shift procedure. Proper selection of the kernel bandwidth is a critical step for a better quality in the clustering. This paper presents a solution for the bandwidth selection, which is completely nonparametric and is based on the sample point estimator to yield a spatial pattern of local bandwidths. The method was applied to synthetic brain images, showing a high performance even in the presence of varying noise level and bias

Automatic initialization for a snakes-based cardiac contour extraction

Segmentation of cardiac structures has a great impact in the quantification of parameters indicative of heart function. The preferred technique for this segmentation is based on edge detection, where deformable models have extensively been used to obtain cardiac cavities contours. This work proposes two alternatives to the traditionally used active model. The first consists of a pre-segmentation algorithm to automatically obtain an initial contour, while the second is a snakes formulation taking into account the explicit analytical expression of the natural cubic splines. The algorithms were tested by comparing their performance in two types of MRI cardiac sequences and the obtained results were compared with the manual tracing of contours as defined by an expert.

Trachea segmentation in CT images using active contours

Tracheal stenosis is an uncommon pathology that in early stages is often confused with different respiratory affections by its signs and symptoms. An automatic characterization of the tracheal stenosis requires adequate medical images and efficient segmentation algorithms. In CT images, several algorithms of airway segmentation have been used, such as 3D region growing, thresholding and gray-level profile analysis. In this work a segmentation method for trachea extraction in CT images is proposed. The algorithm is based on an active contour model (SS) formulated by considering the explicit expression of the natural cubic splines and is compared with the original snakes model (OS). In both cases, an automatic definition of the initial contour based on a Canny filter is proposed. Eight images were processed with both algorithms and the results show that the SS model is less sensitive to initial conditions. For this image modality the Canny operator proved to be a good choice to obtain the initial contour. The SS method generates a smoothed version of the tracheal border.

Texture-based echocardiographic segmentation using a non-parametric estimator and an active contour model

An accurate segmentation of cardiac cavities is necessary to assess cardiac function and to determine quantitative parameters. Several semi-automatic techniques have been tested to achieve this goal. In this work we propose an algorithm to segment cardiac structures, based on a robust pre-processing step that eliminates noise and extracts an initial frontier, together with a refined deformable model, that integrates edge confidence and texture information. Results show that a combination of a mean-shift filter with an active contour model is adequate for echographic images, especially when texture information is included.