Catálogo de publicaciones - libros

In many medical imaging scenarios, real-time high-quality anatomical data visualization and interaction is important to the physician for meaningful diagnosis 3D medical data and get timely feedback. Unfortunately, it is still difficult to achieve an optimized balance between real-time artifact-free medical image volume rendering and interactive data classification. In this paper, we present a new segment-based post color-attenuated classification algorithm to address this problem. In addition, we apply an efficient numerical integration computation technique and take advantage of the symmetric storage format of the color lookup table generation matrix. When implemented within our GPU-based volume raycasting system, the new classification technique is about 100 times faster than the unaccelerated pre-integrated classification approach, while achieving the similar or even superior quality volume rendered image. In addition, we propose an objective measure of artifacts in rendered medical image based on high-frequency spatial image content.

Surgeons need a robust interventional system capable of providing reliable, real-time information regarding the position and orientation of the surgical targets and tools to compensate for the lack of direct vision and to enhance manipulation of intracardiac targets during minimally-invasive, off-pump cardiac interventions. In this paper, we describe a novel method for creating dynamic, pre-operative, subject-specific cardiac models containing the surgical targets and surrounding anatomy, and how they are used to augment the intra-operative virtual environment for guidance of valvular interventions. The accuracy of these pre-operative models was established by comparing the target registration error between the mitral valve annulus characterized in the pre-operative images and their equivalent structures manually extracted from 3D US data. On average, the mitral valve annulus was extracted with a 3.1 mm error across all cardiac phases. In addition, we also propose a method for registering the pre-operative models into the intra-operative virtual environment.

The increasing use of robotic assisted minimally invasive surgery (MIS) provides an ideal environment for using Augmented Reality (AR) for performing image guided surgery. Seamless synthesis of AR depends on a number of factors relating to the way in which virtual objects appear and visually interact with a real environment. Traditional overlaid AR approaches generally suffer from a loss of depth perception. This paper presents a new AR method for robotic assisted MIS, which uses a novel -space based non-photorealistic rendering technique for providing see-through vision of the embedded virtual object whilst maintaining salient anatomical details of the exposed anatomical surface. Experimental results with both phantom and lung lobectomy data demonstrate the visual realism achieved for the proposed method and its accuracy in providing high fidelity AR depth perception.

In today’s climate of clinical governance there is growing pressure on surgeons to demonstrate their competence, improve standards and reduce surgical errors. This paper presents a study on developing a novel eye-gaze driven technique for surgical assessment and workflow recovery. The proposed technique investigates the use of a Parallel Layer Perceptor (PLP) to automate the recognition of a key surgical step in a porcine laparoscopic cholecystectomy model. The classifier is eye-gaze contingent but combined with image based visual feature detection for improved system performance. Experimental results show that by fusing image instrument likelihood measures, an overall classification accuracy of 75% is achieved.

We present a novel automatic multiscale algorithm applied to segmentation of anatomical structures in brain MRI. The algorithm which is derived from algebraic multigrid, uses a graph representation of the image and performs a coarsening process that produces a full hierarchy of segments. Our main contribution is the incorporation of prior knowledge information into the multiscale framework through a Bayesian formulation. The probabilistic information is based on an atlas prior and on a likelihood function estimated from a manually labeled training set. The significance of our new approach is that the constructed pyramid, reflects the prior knowledge formulated. This leads to an accurate and efficient methodology for detection of various anatomical structures simultaneously. Quantitative validation results on gold standard MRI show the benefit of our approach.

Understanding the rapid evolution of cerebral cortical surfaces in developing neonates is essential in order to understand normal human brain development and to study anatomical abnormalities in preterm infants. Several methods to model and align cortical surfaces for cross-sectional studies have been developed. However, the registration of cortical surfaces extracted from neonates across different gestational ages for longitudinal studies remains difficult because of significant cerebral growth. In this paper, we present an automatic cortex registration algorithm, based on surface relaxation followed by non-rigid surface registration. This technique aims to establish the longitudinal spatial correspondence of cerebral cortices for the developing brain in neonates. The algorithm has been tested on 5 neonates. Each infant has been scanned at three different time points. Quantitative results are obtained by propagating sulci across multiple gestational ages and computing the overlap ratios with manually established ground-truth.

This paper presents a discriminative model of multivariate pattern classification, based on functional magnetic resonance imaging (fMRI) and anatomical template. As a measure of brain function, Regional homogeneity (ReHo) is calculated voxel by voxel, and then a widely used anatomical template is applied on ReHo map to parcelate it into 116 brain regions. The mean and standard deviation of ReHo values in each region are extracted as features. Pseudo-Fisher Linear Discriminant Analysis (PFLDA) is performed for training samples to generate discriminative model. Classification experiments have been carried out in 48 schizophrenia patients and 35 normal controls. Under a full leave-one-out (LOO) cross-validation, correct prediction rate of 80% is achieved. Anatomical parcellation process is proved useful to improve classification rate by a control experiment. The discriminative model shows its ability to reveal abnormal brain functional activities and identify people with schizophrenia.

This paper presents a novel and fast probabilistic method for white matter fiber tracking from diffusion weighted MRI (DWI). We formulate fiber tracking on a nonlinear state space model which is able to capture both smoothness regularity of fibers and uncertainties of the local fiber orientations due to noise and partial volume effects. The global tracking model is implemented using particle filtering, which allows us to recursively compute the posterior distribution of the potential fibers. The fiber orientation distribution is theoretically formulated for prolate and oblate tensors separately. Fast and efficient sampling is realised using the von Mises-Fisher distribution on unit spheres. Given a seed point, the method is able to rapidly locate the global optimal fiber and also provide a connectivity map. The proposed method is demonstrated on a brain dataset.

Diffusion tensor magnetic resonance imaging (DT-MRI) based fiber tractography aims at reconstruction of the fiber network of brain. Most commonly employed techniques for fiber tractography are based on the numerical integration of the principal diffusion directions. Although these approaches generate intuitive and easy to interpret results, they are prone to cumulative errors and mostly discard the stochastic nature of DT-MRI data. The proposed Split & Merge Tractography (SMT) technique aims at overcoming the drawbacks of fiber tractography by incorporating it with Markov Chain Monte Carlo techniques. SMT is based on clustering diversely distributed short fiber tracts based on their inter-connectivity. SMT also provides real-time interaction to adjust a user defined confidence level for clustering.

Multisubject statistical analyses of diffusion tensor images in regions of specific white matter tracts have commonly measured only the mean value of a scalar invariant such as the fractional anisotropy (FA), ignoring the spatial variation of FA along the length of fiber tracts. We propose to instead perform tract-based morphometry (TBM), or the statistical analysis of diffusion MRI data in an anatomical tract-based coordinate system. We present a method for automatic generation of white matter tract arc length parameterizations, based on learning a fiber bundle model from tractography from multiple subjects. Our tract-based coordinate system enables TBM for the detection of white matter differences in groups of subjects. We present example TBM results from a study of interhemispheric differences in FA.