Michael Sass Hansen

A path algorithm for the support vector domain description and its application to medical imaging

The support vector domain description is a one-class classification method that estimates the distributional support of a data set. A flexible closed boundary function is used to separate trustworthy data on the inside from outliers on the outside. A single regularization parameter determines the shape of the boundary and the proportion of observations that are regarded as outliers. Picking an appropriate amount of regularization is crucial in most applications but is, for computational reasons, commonly limited to a small collection of parameter values. This paper presents an algorithm where the solutions for all possible values of the regularization parameter are computed at roughly the same computational complexity previously required to obtain a single solution. Such a collection of solutions is known as a regularization path. Knowledge of the entire regularization path not only aids model selection, but may also provide new information about a data set. We illustrate this potential of the method in two applications; one where we establish a sensible ordering among a set of corpora callosa outlines, and one where ischemic segments of the myocardium are detected in patients with acute myocardial infarction.

Deformable Mosaicing for Whole-Body MRI

Whole-body magnetic resonance imaging is an emerging application gaining vast clinical interest during the last years. Although recent technological advances shortened the longish acquisition time, this is still the limiting factor avoiding its wide-spread clinical usage. The acquisition of images with large field-of-view helps to relieve this drawback, but leads to significantly distorted images. Therefore, we propose a deformable mosaicing approach, based on the simultaneous registration to linear weighted averages, to correct for distortions in the overlapping area. This method produces good results on in-vivo data and has the advantage that a seamless integration into the clinical workflow is possible.

Robust pseudo-hierarchical support vector clustering

Support vector clustering (SVC) has proven an efficient algorithm for clustering of noisy and high-dimensional data sets, with applications within many fields of research. An inherent problem, however, has been setting the parameters of the SVC algorithm. Using the recent emergence of a method for calculating the entire regularization path of the support vector domain description, we propose a fast method for robust pseudo-hierarchical support vector clustering (HSVC). The method is demonstrated to work well on generated data, as well as for detecting ischemic segments from multidimensional myocardial perfusion magnetic resonance imaging data, giving robust results while drastically reducing the need for parameter estimation.

Sparse statistical deformation model for the analysis of craniofacial malformations in the Crouzon mouse

Crouzon syndrome is characterised by the premature fusion of cranial sutures. Recently the first genetic Crouzon mouse model was generated. In this study, Micro CT skull scannings of wild-type mice and Crouzon mice were investigated. Using nonrigid registration, a wild-type craniofacial mouse atlas was built. The atlas was registered to all mice providing parameters controlling the deformations for each subject. Our previous PCA-based statistical deformation model on these parameters revealed only one discriminating mode of variation. Aiming at distributing the discriminating variation over more modes we built a different model using Independent Component Analysis (ICA). Here, we focus on a third method, sparse PCA (SPCA), which aims at approximating the properties of a standard PCA while introducing sparse modes of variation. The results show that SPCA outperforms both ICA and PCA with respect to the Fisher discriminant, although many similarities are found with respect to ICA.

Ischemic segment detection using the support vector domain description

Myocardial perfusion Magnetic Resonance (MR) imaging has proven to be a powerful method to assess coronary artery diseases. The current work presents a novel approach to the analysis of registered sequences of myocardial perfusion MR images. A previously reported active appearance model (AAM) based segmentation and registration of the myocardium provided pixel-wise signal intensity curves that were analyzed using the Support Vector Domain Description (SVDD). In contrast to normal SVDD, the entire regularization path was calculated and used to calculate a generalized distance, which is used to discriminate between ischemic and healthy tissue. The results corresponded well to the ischemic segments found by assessment of the three common perfusion parameters; maximum upslope, peak and time-to-peak obtained pixel-wise.

Registration-based interpolation applied to cardiac MRI

Various approaches have been proposed for segmentation of cardiac MRI. An accurate segmentation of the myocardium and ventricles is essential to determine parameters of interest for the function of the heart, such as the ejection fraction. One problem with MRI is the poor resolution in one dimension. A 3D registration algorithm will typically use a trilinear interpolation of intensities to determine the intensity of a deformed template image. Due to the poor resolution across slices, such linear approximation is highly inaccurate since the assumption of smooth underlying intensities is violated. Registration-based interpolation is based on 2D registrations between adjacent slices and is independent of segmentations. Hence, rather than assuming smoothness in intensity, the assumption is that the anatomy is consistent across slices. The basis for the proposed approach is the set of 2D registrations between each pair of slices, both ways. The intensity of a new slice is then weighted by (i) the deformation functions and (ii) the intensities in the warped images. Unlike the approach by Penney et al. 2004, this approach takes into account deformation both ways, which gives more robustness where correspondence between slices is poor. We demonstrate the approach on a toy example and on a set of cardiac CINE MRI. Qualitative inspection reveals that the proposed approach provides a more convincing transition between slices than images obtained by linear interpolation. A quantitative validation reveals significantly lower reconstruction errors than both linear and registration-based interpolation based on one-way registrations.

Computing minimal deformations: application to construction of statistical shape models

Nonlinear registration is mostly performed after initialization by a global, linear transformation (in this work, we focus on similarity transformations), computed by a linear registration method. For the further processing of the results, it is mostly assumed that this preregistration step completely removes the respective linear transformation. However, we show that in deformable settings, this is not the case. As a consequence, a significant linear component is still existent in the deformation computed by the nonlinear registration algorithm. For construction of statistical shape models (SSM) from deformations, this is an unwanted property: SSMs should not contain similarity transformations, since these do not capture information about shape. We propose a method which performs an a posteriori extraction of a similarity transformation from a given nonlinear deformation field, and we use the processed fields as input for SSM construction. For computation of minimal displacements, a closed-form solution minimizing the squared Euclidean norm of the displacement field subject to similarity parameters is used. Experiments on real inter-subject data and on a synthetic example show that the theoretically justified removal of the similarity component by the proposed method has a large influence on the shape model and significantly improves the results.

Detection of connective tissue disorders from 3d aortic MR images using independent component analysis

A computer-aided diagnosis (CAD) method is reported that allows the objective identification of subjects with connective tissue disorders from 3D aortic MR images using segmentation and independent component analysis (ICA). The first step to extend the model to 4D (3D + time) has also been taken. ICA is an effective tool for connective tissue disease detection in the presence of sparse data using prior knowledge to order the components, and the components can be inspected visually. n n3D+time MR image data sets acquired from 31 normal and connective tissue disorder subjects at end-diastole (R-wave peak) and at 45% of the R-R interval were used to evaluate the performance of our method. The automated 3D segmentation result produced accurate aortic surfaces covering the aorta. The CAD method distinguished between normal and connective tissue disorder subjects with a classification accuracy of 93.5 %.

ESTIMATION OF INDEPENDENT NON-LINEAR DEFORMATION MODES FOR ANALYSIS OF CRANIOFACIAL MALFORMATIONS IN CROUZON MICE

Crouzon syndrome is a genetic disease resulting in premature fusion of cranial sutures and synchondroses causing craniosynostosis. A decade ago the Crouzon gene was discovered, and recently the first mouse model of the syndrome was generated. In this study, a set of micro CT scannings of the heads of wild-type (normal) mice and Crouzon mice were investigated. We present for what we believe is the first time, a statistical deformation model based on independent component analysis (ICA). A set of deformation parameters for each mouse was calculated using a B-spline-based non-rigid registration. From the parameters controlling the deformations for each subject, the statistical model was estimated. ICA is demonstrated to provide localized deformation components, many of which give a clear separation between Crouzon and wild-type mice. This is a clear improvement of a previous principal component-based model, which only provided one global deformation component describing the disease. The ICA components allow interpretation of each deformation feature to be carried out independently of other features, and provides a basis for linking the observed craniofacial malformations to the fusing of sutures. ICA revealed an interesting new finding, not previously reported in the literature, namely asymmetries in the head in Crouzon mice. This phenomenon is probably caused by asymmetric closure of craniofacial sutures

Registration-based interpolation real-time volume visualization

Rendering tomographic data sets is a computationally expensive task, and often accomplished using hardware acceleration. The data sets are usually anisotropic as a result of the process used to acquire them. A vital part of rendering them is the conversion of the discrete signal back into a continuous one, via interpolation. On graphics hardware, this is often achieved via simple linear interpolation.n We present a novel approach for real-time anisotropic volume data interpolation on a graphics processing unit and draw comparisons to standardized interpolation alternatives. Our approach uses a pre-computed set of cross-slice correspondences to compensate for missing data. We perform a qualitative analysis using sparse data sets, investigating both visual quality, as well divergence from the ground truth, testing the limits of the interpolation method.n Our method produces high quality interpolation with a moderate performance impact compared to alternatives. It is ideal for reconstructing sparse data sets, as well as minimizing quality loss while scaling large amounts of data to fit on most mobile graphics cards.