| 1. |
- Borga, Magnus, 1965-
(author)
-
Learning Multidimensional Signal Processing
- 1998
-
Doctoral thesis (other academic/artistic)abstract
- The subject of this dissertation is to show how learning can be used for multidimensional signal processing, in particular computer vision. Learning is a wide concept, but it can generally be defined as a system’s change of behaviour in order to improve its performance in some sense.Learning systems can be divided into three classes: supervised learning, reinforcement learning and unsupervised learning. Supervised learning requires a set of training data with correct answers and can be seen as a kind of function approximation. A reinforcement learning system does not require a set of answers. It learns by maximizing a scalar feedback signal indicating the system’s performance. Unsupervised learning can be seen as a way of finding a good representation of the input signals according to a given criterion.In learning and signal processing, the choice of signal representation is a central issue. For high-dimensional signals, dimensionality reduction is often necessary. It is then important not to discard useful information. For this reason, learning methods based on maximizing mutual information are particularly interesting.A properly chosen data representation allows local linear models to be used in learning systems. Such models have the advantage of having a small number of parameters and can for this reason be estimated by using relatively few samples. An interesting method that can be used to estimate local linear models is canonical correlation analysis (CCA). CCA is strongly related to mutual information. The relation between CCA and three other linear methods is discussed. These methods are principal component analysis (PCA), partial least squares (PLS) and multivariate linear regression (MLR). An iterative method for CCA, PCA, PLS and MLR, in particular low-rank versions of these methods, is presented.A novel method for learning filters for multidimensional signal processing using CCA is presented. By showing the system signals in pairs, the filters can be adapted to detect certain features and to be invariant to others. A new method for local orientation estimation has been developed using this principle. This method is significantly less sensitive to noise than previously used methods.Finally, a novel stereo algorithm is presented. This algorithm uses CCA and phase analysis to detect the disparity in stereo images. The algorithm adapts filters in each local neighbourhood of the image in a way which maximizes the correlation between the filtered images. The adapted filters are then analysed to find the disparity. This is done by a simple phase analysis of the scalar product of the filters. The algorithm can even handle cases where the images have different scales. The algorithm can also handle depth discontinuities and give multiple depth estimates for semi-transparent images.
|
|
| 2. |
- Cros, Olivier
(author)
-
Structural properties of the mastoid using image analysis and visualization
- 2017
-
Doctoral thesis (other academic/artistic)abstract
- The mastoid, located in the temporal bone, houses an air cell system whose cells have a variation in size that can go far below current conventional clinical CT scanner resolution. Therefore, the mastoid air cell system is only partially represented in a CT scan. Where the conventional clinical CT scanner lacks level of minute details, micro-CT scanning provides an overwhelming amount of ne details. The temporal bone being one of the most complex in the human body, visualization of micro-CT scanning of this boneawakens the curiosity of the experimenter, especially with the correct visualization settings.This thesis first presents a statistical analysis determining the surface area to volume ratio of the mastoid air cell system of human temporal bone, from micro-CT scanning using methods previously applied for conventional clinical CT scans. The study compared current results with previous studies, with successive downsampling the data down to a resolution found in conventional clinical CT scanning. The results from the statistical analysis showed that all the small mastoid air cells, that cannot be detected in conventional clinical CT scans, do heavily contribute to the estimation of the surface area, and in consequence to the estimation of the surface area to volume ratio by a factor of about 2.6. Such a result further strengthens the idea of the mastoid to play an active role in pressure regulation and gas exchange.Discovery of micro-channels through specific use of a non-traditional transfer function was then reported, where a qualitative and a quantitative pre-analysis were performed and reported. To gain more knowledge about these micro-channels, a local structure tensor analysis was applied where structures are described in terms of planar, tubular, or isotropic structures. The results from this structural tensor analysis suggest these microchannels to potentially be part of a more complex framework, which hypothetically would provide a separate blood supply for the mucosa lining the mastoid air cell system.The knowledge gained from analysing the micro-channels as locally providing blood to the mucosa, led to the consideration of how inflammation of the mucosa could impact the pneumatization of the mastoid air cell system. Though very primitive, a 3D shape analysis of the mastoid air cell system was carried out. The mastoid air cell system was first represented in a compact form through a medial axis, from which medial balls could be used. The medial balls, representative of how large the mastoid air cells can be locally, were used in two complementary clustering methods, one based on the size diameter of the medial balls and one based on their location within the mastoid air cell system. From both quantitative and qualitative statistics, it was possible to map the clusters based on pre-defined regions already described in the literature, which opened the door for new hypotheses concerning the effect of mucosal inflammation on the mastoid pneumatization.Last but not least, discovery of other structures, previously unreported in the literature, were also visually observed and briefly discussed in this thesis. Further analysis of these unknown structures is needed.
|
|
| 3. |
- Eklund, Anders, 1981-
(author)
-
Computational Medical Image Analysis : With a Focus on Real-Time fMRI and Non-Parametric Statistics
- 2012
-
Doctoral thesis (other academic/artistic)abstract
- Functional magnetic resonance imaging (fMRI) is a prime example of multi-disciplinary research. Without the beautiful physics of MRI, there wouldnot be any images to look at in the first place. To obtain images of goodquality, it is necessary to fully understand the concepts of the frequencydomain. The analysis of fMRI data requires understanding of signal pro-cessing, statistics and knowledge about the anatomy and function of thehuman brain. The resulting brain activity maps are used by physicians,neurologists, psychologists and behaviourists, in order to plan surgery andto increase their understanding of how the brain works.This thesis presents methods for real-time fMRI and non-parametric fMRIanalysis. Real-time fMRI places high demands on the signal processing,as all the calculations have to be made in real-time in complex situations.Real-time fMRI can, for example, be used for interactive brain mapping.Another possibility is to change the stimulus that is given to the subject, inreal-time, such that the brain and the computer can work together to solvea given task, yielding a brain computer interface (BCI). Non-parametricfMRI analysis, for example, concerns the problem of calculating signifi-cance thresholds and p-values for test statistics without a parametric nulldistribution.Two BCIs are presented in this thesis. In the first BCI, the subject wasable to balance a virtual inverted pendulum by thinking of activating theleft or right hand or resting. In the second BCI, the subject in the MRscanner was able to communicate with a person outside the MR scanner,through a virtual keyboard.A graphics processing unit (GPU) implementation of a random permuta-tion test for single subject fMRI analysis is also presented. The randompermutation test is used to calculate significance thresholds and p-values forfMRI analysis by canonical correlation analysis (CCA), and to investigatethe correctness of standard parametric approaches. The random permuta-tion test was verified by using 10 000 noise datasets and 1484 resting statefMRI datasets. The random permutation test is also used for a non-localCCA approach to fMRI analysis.
|
|
| 4. |
- Landelius, Tomas
(author)
-
Reinforcement Learning and Distributed Local Model Synthesis
- 1997
-
Doctoral thesis (other academic/artistic)abstract
- Reinforcement learning is a general and powerful way to formulate complex learning problems and acquire good system behaviour. The goal of a reinforcement learning system is to maximize a long term sum of instantaneous rewards provided by a teacher. In its extremum form, reinforcement learning only requires that the teacher can provide a measure of success. This formulation does not require a training set with correct responses, and allows the system to become better than its teacher.In reinforcement learning much of the burden is moved from the teacher to the training algorithm. The exact and general algorithms that exist for these problems are based on dynamic programming (DP), and have a computational complexity that grows exponentially with the dimensionality of the state space. These algorithms can only be applied to real world problems if an efficient encoding of the state space can be found.To cope with these problems, heuristic algorithms and function approximation need to be incorporated. In this thesis it is argued that local models have the potential to help solving problems in high-dimensional spaces and that global models have not. This is motivated with the biasvariance dilemma, which is resolved with the assumption that the system is constrained to live on a low-dimensional manifold in the space of inputs and outputs. This observation leads to the introduction of bias in terms of continuity and locality.A linear approximation of the system dynamics and a quadratic function describing the long term reward are suggested to constitute a suitable local model. For problems involving one such model, i.e. linear quadratic regulation problems, novel convergence proofs for heuristic DP algorithms are presented. This is one of few available convergence proofs for reinforcement learning in continuous state spaces.Reinforcement learning is closely related to optimal control, where local models are commonly used. Relations to present methods are investigated, e.g. adaptive control, gain scheduling, fuzzy control, and jump linear systems. Ideas from these areas are compiled in a synergistic way to produce a new algorithm for heuristic dynamic programming where function parameters and locality, expressed as model applicability, are learned on-line. Both top-down and bottom-up versions are presented.The emerging local models and their applicability need to be memorized by the learning system. The binary tree is put forward as a suitable data structure for on-line storage and retrieval of these functions.
|
|
| 5. |
- Sigfridsson, Andreas, 1978-
(author)
-
Multidimensional MRI of Myocardial Dynamics : Acquisition, Reconstruction and Visualization
- 2009
-
Doctoral thesis (other academic/artistic)abstract
- Methods for measuring deformation and motion of the human heart in-vivo are crucial in the assessment of cardiac function. Applications ranging from basic physiological research, through early detection of disease to follow-up studies, all rely on the quality of the measurements of heart dynamics. This thesis presents new improved magnetic resonance imaging methods for acquisition, image reconstruction and visualization of cardiac motion and deformation.As the heart moves and changes shape during the acquisition, synchronization to the heart dynamics is necessary. Here, a method to resolve not only the cardiac cycle but also the respiratory cycle is presented. Combined with volumetric imaging, this produces a five-dimensional data set with two cyclic temporal dimensions. This type of data reveals unique physiological information, such as interventricular coupling in the heart in different phases of the respiratory cycle.The acquisition can also be sensitized to motion, measuring not only the magnitude of the magnetization but also a signal proportional to local velocity or displacement. This allows for quantification of the motion which is especially suitable for functional study of the cardiac deformation. In this work, an evaluation of the influence of several factors on the signal-to-noise ratio is presented for in-vivo displacement encoded imaging. Additionally, an extension of the method to acquire multiple displacement encoded slices in a single breath hold is also presented.Magnetic resonance imaging is usually associated with long scan times, and many methods exist to shorten the acquisition time while maintaining acceptable image quality. One class of such methods involves acquiring only a sparse subset of k-space. A special reconstruction is then necessary in order to obtain an artifact-free image. One family of these reconstruction techniques tailored for dynamic imaging is the k-t BLAST approach, which incorporates data-driven prior knowledge to suppress aliasing artifacts that otherwise occur with the sparse sampling. In this work, an extension of the original k-t BLAST method to two temporal dimensions is presented and applied to data acquired with full coverage of the cardio-respiratory cycles. Using this technique, termed k-t2 BLAST, simultaneous reduction of scan time and improved spatial resolution is demonstrated. Further, the loss of temporal fidelity when using the k-t BLAST approach is investigated, and an improved reconstruction is proposed for the application of cardiac function analysis.Visualization is a crucial part of the imaging chain. Scalar data, such as regular anatomical images, are straightforward to display. Myocardial strain and strain-rate, however, are tensor quantities which do not lend themselves to direct visualization. The problem of visualizing the tensor field is approached in this work by combining a local visualization that displays all degrees of freedom for a single tensor with an overview visualization using a scalar field representation of the complete tensor field. The scalar field is obtained by iterated adaptive filtering of a noise field, creating a continuous geometrical representation of the myocardial strain-rate tensor field.The results of the work presented in this thesis provide opportunities for improved imaging of myocardial function, in all areas of the imaging chain; acquisition, reconstruction and visualization.
|
|
| 6. |
- Zikrin, Spartak, 1987-
(author)
-
Large-Scale Optimization Methods with Application to Design of Filter Networks
- 2014
-
Doctoral thesis (other academic/artistic)abstract
- Nowadays, large-scale optimization problems are among those most challenging. Any progress in developing methods for large-scale optimization results in solving important applied problems more effectively. Limited memory methods and trust-region methods represent two ecient approaches used for solving unconstrained optimization problems. A straightforward combination of them deteriorates the efficiency of the former approach, especially in the case of large-scale problems. For this reason, the limited memory methods are usually combined with a line search. We develop new limited memory trust-region algorithms for large-scale unconstrained optimization. They are competitive with the traditional limited memory line-search algorithms.In this thesis, we consider applied optimization problems originating from the design of lter networks. Filter networks represent an ecient tool in medical image processing. It is based on replacing a set of dense multidimensional lters by a network of smaller sparse lters called sub-filters. This allows for improving image processing time, while maintaining image quality and the robustness of image processing.Design of lter networks is a nontrivial procedure that involves three steps: 1) choosing the network structure, 2) choosing the sparsity pattern of each sub-filter and 3) optimizing the nonzero coecient values. So far, steps 1 and 2 were mainly based on the individual expertise of network designers and their intuition. Given a sparsity pattern, the choice of the coecients at stage 3 is related to solving a weighted nonlinear least-squares problem. Even in the case of sequentially connected lters, the resulting problem is of a multilinear least-squares (MLLS) type, which is a non-convex large-scale optimization problem. This is a very dicult global optimization problem that may have a large number of local minima, and each of them is singular and non-isolated. It is characterized by a large number of decision variables, especially for 3D and 4D lters.We develop an effective global optimization approach to solving the MLLS problem that reduces signicantly the computational time. Furthermore, we develop efficient methods for optimizing sparsity of individual sub-filters in lter networks of a more general structure. This approach offers practitioners a means of nding a proper trade-o between the image processing quality and time. It allows also for improving the network structure, which makes automated some stages of designing lter networks.
|
|
