Raquel Valdés-Cristerna

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.

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

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.

Active contours and surfaces with cubic splines for semiautomatic tracheal segmentation

Signs and symptoms of tracheal stenosis can create confusion about the etiology of the problem. While bronchoscopy is the diagnostic method of choice to evaluate the extension and localization of the lesion, the use of x-ray computed axial tomography (CAT) images has also been considered. Recent works on airway segmentation in CAT images propose the extensive use of automatic segmentation techniques based on 3-D region growing. This technique is computationally expensive and thus alternative analysis procedures are still under development. We present a segmentation method constructed over an active surface model based on cubic splines interpolation. The 3-D rendering of the upper-airway path segmented from neck and thorax CAT scans using the proposed method is validated in regard to its possible use as a diagnostic tool for the characterization of tracheal stenosis. The results presented relative to the performance of the model, both on synthetic and real CAT scan volumes, indicate that the proposed procedure improves over the reference active model methods.

Factorial phase analysis of ventricular contraction using equilibrium radionuclide angiography images

Abstract Factorial phase analysis (FaPI) represents an alternative method to Fourier phase analysis (FoPI) in the evaluation and detection of abnormalities on cardiac contraction patterns, but it has limitations in representing the sequence in abnormal contraction patterns. In this work we propose a modified factorial phase image (FaPIm) that incorporates more complete information regarding the ventricular contraction sequence. In particular, we analyze and evaluate the contribution of the third eigenimage, in the presence of ventricular dyssynchrony, which has not been sufficiently explored in the literature. We have validated the proposed FaPIm using two Equilibrium Radionuclide Angiography (ERNA) sets of images obtained with a dynamic cardiac phantom and with a numerically simulated phantom. Also, we have tested the proposed representation for a control group of 23 normal subjects and for a sample of 15 patients with Complete Left Bundle Branch Block (LBBB). Whereas FoPI allows us to obtain an image that synthesizes ventricular contraction with the smallest dispersion around the mean values, FaPI and FaPIm show that external areas surrounding ventricular cavities present more dephasing than the rest of the ventricular region and contain more detailed information about the progression of contraction. Also, in the presence of an abnormal contraction pattern, the magnitude of the third eigenvalue was greater than the corresponding eigenvalue obtained for normal simulations. The dispersion plots obtained for a normal contraction pattern show that left ventricle (LV) and right ventricle (RV) information overlap. Therefore, when there is a dyssynchrony between LV and RV contraction it becomes necessary to incorporate the information corresponding to the third factor to achieve a clear separation between regions. In the comparison of the indices of control and LBBB populations, FaPIm shows significant differences in five out of six contraction indices, showing its promising value as a clinical tool.

RBF network with cylindrical coordinate features for multispectral MRI segmentation

Spatial quantification of relevant brain structures, is usually carried out through the analysis of a stack of magnetic resonance (MR) images by means of some image segmentation approach. In this paper, multispectral MR imaging segmentation based on a modified radial-basis function network is presented. Multispectral MR image sets are constructed by collecting data for the same anatomical structures under T1, T2 and FLAIR excitation sequences. Classification features for the network are extended beyond the normalized intensities in each band to also include the cylindrical coordinates of the image pixels. Such coordinates are determined within a reference image space upon which all targets are registered to. The network classifier was designed to differentiate three structures: gray matter, white matter and image background. The classification layer was also modified to accommodate the pixel cylindrical coordinates as inputs. With the designed network, background pixels are correctly classified for all cases, while gray and white matter pixels are misclassified for about 10% of the cases in the validation set. The source of these errors can be traced to smooth transitions in the output nodes for these two classes. Thresholding the outputs of these nodes to include a reject class reduces the misclassification error. The small and simple architecture of the network shows good generalization, and thus good segmentation over unseen stacks.

Factor analysis of ventricular contraction using SPECT-ERNA images

Equilibrium radionuclide angiography images (ERNA) has been established as a useful modality for clinical evaluation of the ventricular function. Tomographic acquisition of ERNA (SPECT-ERNA) improves the quantification of ventricular function with planar ERNA, avoiding both the overlap of structures and the need of defining the best septal view which can be difficult in dilated ventricles. In this work we analyze the contribution and distribution of the most significant factors of dynamic structures (FADS), and propose an index based on the characterization of the normal contraction pattern, to quantify the ventricular contraction normality in a set of patients with clinical diagnosis of pulmonary arterial hypertension (PAH) using SPECT-ERNA. The statistical analysis shows significant differences between normal and PAH subjects in the models of left ventricle (LV) contraction pattern. This comparison shows that the LV has an abnormal contraction as a consequence of the pulmonary arterial hypertension.

Adaptive RBF network with active contour coupling for multispectral MRI segmentation

A segmentation procedure using a radial basis function network (RBFN), coupled with an active contour (AC) model based on a cubic splines formulation is presented for the detection of the gray-white matter boundary in axial MMRI (T1, T2 and PD). A RBFN classifier has been previously introduced for MMRI segmentation, with good generalization at a rate of 10% misclassification over white and gray matter pixels on the validation set. The coupled RBFN and AC model system incorporates the posterior probability estimation map into the AC energy term as a restriction force. The RBFN output is also employed to provide an initial contour for the AC. Furthermore, an adaptation strategy for the network weights, guided by a feedback from the contour model adjustment at each iteration, is described. In order to compare the algorithms performance, the segmentations using the adaptive, as well as the non-adaptive schemes were computed. It was observed that the major differences are located around deep circonvolutions, where the result of the adaptive process is superior than that obtained with the non-adaptive scheme, even in moderate noise conditions. In summary, the RBFN provides a good initial contour for the AC, the coupling of both processes keeps the final contour within the desired region and the adaptive strategy enhances the contour location.

Automated Classification of Severity in Cardiac Dyssynchrony Merging Clinical Data and Mechanical Descriptors

Cardiac resynchronization therapy (CRT) improves functional classification among patients with left ventricle malfunction and ventricular electric conduction disorders. However, a high percentage of subjects under CRT (20%–30%) do not show any improvement. Nonetheless the presence of mechanical contraction dyssynchrony in ventricles has been proposed as an indicator of CRT response. This work proposes an automated classification model of severity in ventricular contraction dyssynchrony. The model includes clinical data such as left ventricular ejection fraction (LVEF), QRS and P-R intervals, and the 3 most significant factors extracted from the factor analysis of dynamic structures applied to a set of equilibrium radionuclide angiography images representing the mechanical behavior of cardiac contraction. A control group of 33 normal volunteers (28 ± 5 years, LVEF of 59.7% ± 5.8%) and a HF group of 42 subjects (53.12 ± 15.05 years, LVEF < 35%) were studied. The proposed classifiers had hit rates of 90%, 50%, and 80% to distinguish between absent, mild, and moderate-severe interventricular dyssynchrony, respectively. For intraventricular dyssynchrony, hit rates of 100%, 50%, and 90% were observed distinguishing between absent, mild, and moderate-severe, respectively. These results seem promising in using this automated method for clinical follow-up of patients undergoing CRT.

P300-based brain-computer interfaces

Abstract Diverse neuro-psychological tasks have been designed for expressing intention mostly in the form of evoked potentials uncovered by time-coherent averaging, spectral analysis, or other digital signal processing techniques. Among others, the elicitation of the so-called P300 event-related potential has been found reproducible, reliable and suitable for environment control and communication applications that are not time-critical. The nature, elicitation and exploitation of the P300 potential for BCI is the subject of this chapter, organized in three sections: P300 event-related potential, P300 for spelling task and P300 for non-spelling tasks.