Publikationer

The validity of parametric functional magnetic resonance imaging (fMRI) analysis has only been reported for simulated data.Recent advances in computer science and data sharing make it possible to analyze large amounts of real fMRI data. In this study,1484 rest datasets have been analyzed in SPM8, to estimate true familywise error rates. For a familywise significance threshold of5%, significant activity was found in 1% - 70% of the 1484 rest datasets, depending on repetition time, paradigm and parametersettings. This means that parametric significance thresholds in SPM both can be conservative or very liberal. The main reason forthe high familywise error rates seems to be that the global AR(1) auto correlation correction in SPM fails to model the spectra ofthe residuals, especially for short repetition times. The findings that are reported in this study cannot be generalized to parametricfMRI analysis in general, other software packages may give different results. By using the computational power of the graphicsprocessing unit (GPU), the 1484 rest datasets were also analyzed with a random permutation test. Significant activity was thenfound in 1% - 19% of the datasets. These findings speak to the need for a better model of temporal correlations in fMRI timeseries.

Parametric statistical methods, such as Z-, t-, and F-values are traditionally employed in functional magnetic resonance imaging (fMRI) for identifying areas in the brain that are active with a certain degree of statistical significance. These parametric methods, however, have two major drawbacks. First, it is assumed that the observed data are Gaussian distributed and independent; assumptions that generally are not valid for fMRI data. Second, the statistical test distribution can be derived theoretically only for very simple linear detection statistics. With non-parametric statistical methods, the two limitations described above can be overcome. The major drawback of non-parametric methods is the computational burden with processing times ranging from hours to days, which so far have made them impractical for routine use in single subject fMRI analysis. In this work, it is shown how the computational power of cost-efficient Graphics Processing Units (GPUs) can be used to speed up random permutation tests. A test with 10 000 permutations takes less than a minute, making statistical analysis of advanced detection methods in fMRI practically feasible. To exemplify the permutation based approach, brain activity maps generated by the General Linear Model (GLM) and Canonical Correlation Analysis (CCA) are compared at the same significance level. During the development of the routines and writing of the paper, 3-4 years of processing time has been saved by using the GPU.

Medical image registration is becoming a more and more useful component of a large number of applications. The presented method aims to enrich the ITK library. This method, called Morphon registration algorithm, computes a dense deformation field accepting inputs from different intensity contrasts. This article presents its implementation within the Insight Toolkit. In this paper, we provide a brief description of the algorithm, a presentation of the implementation, the justification of our modified classes and the results given by the algorithm. We demonstrate the algorithm in application of different images intesity constrasts and dimensions.

This paper presents new methods for use of dense motion fields for motion compensation of interlaced video. The motion estimation is based on previously decoded field-images. The motion is then temporally predicted and used for motion compensated prediction of the field-image to be coded. The motion estimation algorithm is phase-based and uses two or three field-images to achieve motion estimates with sub-pixel accuracy. To handle non-constant motion and the specific characteristics of the field-image to be coded, the initially predicted image is refined using forward motion compensation, based on block-matching. Tests show that this approach achieves higher PSNR than forward block-based motion estimation, when coding the residual with the same coder. The subjective performance is also better.

As welfare diseases become more common all over the world the demand for angiography examinations is increasing rapidly. The development of advanced medical signal processing methods has with few exceptions been concentrated towards CT and MR while traditional contrast based radiology depend on methods developed for ancient photography techniques despite the fact that angiography sequences are generally recorded in digital form. This article presents a new approach for processing of angiography sequences based on advanced image processing methods. The developed algorithm automatically processes angiography sequences containing motion artifacts that cannot be processed by conventional methods like digital subtraction angiography (DSA) and pixel shift due to non uniform motions. The algorithm can in simple terms be described as an ideal pixelshift filter carrying out shifts of different directions and magnitude according to the local motions in the image. In difference to conventional methods it is fully automatic, no mask image needs to be defined and the manual pixelshift operations, which are extremely time consuming, are eliminated. The algorithm is efficient and robust and is designed to run on standard hardware of a powerful workstation which excludes the need for expensive dedicated angiography platforms. Since there is no need to make additional recordings if the patient moves, the patient is exposed to less amount of radiation and contrast fluid. The most exciting benefits by this method are, however, that it opens up new areas for contrast based angiography that are not possible to process with conventional methods e.g. nonuniform motions and multiple layers of moving tissue. Advanced image processing methods provide significantly better image quality and noise suppression but do also provide the means to compute flow velocity and visualize the flow dynamics in the arterial trees by e.g. using color. Initial tests have proven that it is possible to discriminate capillary blood flow from angiography data which opens up interesting possibilities for estimating the blood flow in the heart muscle without use of nuclear methods.

Tensor valued data are frequently used in medical imaging. For a  3-dimensional second order tensor such data imply at least six  degrees of freedom for each voxel. The operators ability to perceive  this information is of outmost importance and in many cases a  limiting factor for the interpretation of the data. In this paper we  propose a decomposition of such tensor fields using the Tflash  tensor glyphs that intuitively conveys important tensor features to  a human observer. A matlab implementation for visualization of  single tensors are described in detail and a VTK/ITK implementation  for visualization of tensor fields have been developed as a Medical  Studio component.

Image registration is a well-known concept within the medical image domain and has been shown to be useful in a number of dierent tasks. However, due to sometimes long processing times, image registration is not fully utilized in clinical workows, where time is an important factor. During the last couple of years, a number of signicant projects have been introduced to make the computational power of GPUs available to a wider audience, where the most well known is CUDA. In this paper we present, with the aid of CUDA, a speedup in the range of 38-44x (from 29 minutes to 40 seconds) when implementing a phasebased non-rigid image registration algorithm, known as the Morphon, on a single GPU. The achieved speedup is in the same magnitude as the speedups reported from other non-rigid registration algorithms fully ported to the GPU. Given the impressive speedups, both reported in this paper and other papers, we therefore consider that it is now feasible to eectively integrate image registration into various clinical workows, where time is a critical factor.

The visualization of images with a large dynamic range is a difficult task and this is especially the case for gray-level images. In radiology departments, this will force radiologists to review medical images several times, since the images need to be visualized with several different contrast windows (transfer functions) in order for the full information content of each image to be seen. Previously suggested methods for handling this situation include various approaches using histogram equalization and other methods for processing the image data. However, none of these utilize the underlying human anatomy in the images to control the visualization and the fact that different transfer functions are often only relevant for disjoint anatomical regions. In this paper, we propose a method for using model-based local transfer functions. It allows the reviewing radiologist to apply multiple transfer functions simultaneously to a medical image volume. This provides the radiologist with a tool for making the review process more efficient, by allowing him/her to review more of the information in a medical image volume with a single visualization. The transfer functions are automatically assigned to different anatomically relevant regions, based upon a model registered to the volume to be visualized. The transfer functions can be either pre-defined or interactively changed by the radiologist during the review process. All of this is achieved without adding any unfamiliar aspects to the radiologist’s normal work-flow, when reviewing medical image volumes.

Functional connectivity analysis is a way to investigate how different parts of the brain are connected and interact. A common measure of connectivity is the temporal correlation between a reference voxel time series and all the other time series in a functional MRI data set. An fMRI data set generally contains more than 20,000 within-brain voxels, making a complete correlation analysis between all possible combinations of voxels heavy to compute, store, visualize and explore. In this paper, a GPU-accelerated interactive tool for investigating functional connectivity in fMRI data is presented. A reference voxel can be moved by the user and the correlations to all other voxels are calculated in real-time using the graphics processing unit (GPU). The resulting correlation map is updated in real-time and visualized as a 3D volume rendering together with a high resolution anatomical volume. This tool greatly facilitates the search for interesting connectivity patterns in the brain.

Parametric statistical methods are traditionally employed in functional magnetic resonance imaging (fMRI) for identifying areas in the brain that are active with a certain degree of statistical significance. These parametric methods, however, have two major drawbacks. First, it isassumed that the observed data are Gaussian distributed and independent; assumptions that generally are not valid for fMRI data. Second, the statistical test distribution can be derived theoretically only for very simple linear detection statistics. In this work it is shown how the computational power of the Graphics Processing Unit (GPU) can be used to speedup non-parametric tests, such as random permutation tests. With random permutation tests it is possible to calculate significance thresholds for any test statistics. As an example, fMRI activity maps from the General Linear Model (GLM) and Canonical Correlation Analysis (CCA) are compared at the same significance level.

The aim of this medical image science project is to increase patient safety in terms of improved image quality and reduced exposure to ionizing radiation in CT. The means to achieve these goals is to develop and evaluate an efficient adaptive filtering (denoising/image enhancement) method that fully explores true 4D image acquisition modes. Four-dimensional (4D) medical image data are captured as a time sequence of image volumes. During 4D image acquisition, a 3D image of the patient is recorded at regular time intervals. The resulting data will consequently have three spatial dimensions and one temporal dimension. Increasing the dimensionality of the data impose a major increase the computational demands. The initial linear filtering which is the cornerstone in all adaptive image enhancement algorithms increase exponentially with the dimensionality. On the other hand the potential gain in Signal to Noise Ratio (SNR) also increase exponentially with the dimensionality. This means that the same gain in noise reduction that can be attained by performing the adaptive filtering in 3D as opposed to 2D can be expected to occur once more by moving from 3D to 4D. The initial tests on on both synthetic and clinical 4D images has resulted in a significant reduction of the noise level and an increased detail compared to 2D and 3D methods. When tuning the parameters for adaptive filtering is extremely important to attain maximal diagnostic value which not necessarily coincide with an an eye pleasing image for a layman. Although this application focus on CT the resulting adaptive filtering methods will be beneficial for a wide range of 3D/4D medical imaging modalities e.g. shorter acquisition time in MRI and improved elimination of noise in 3D or 4D ultrasound datasets.

We present a novel approach to interactive and concurrent volume visualization of functional Magnetic Resonance Imaging (fMRI). While the patient is in the scanner, data is extracted in real-time using state-of-the-art signal processing techniques. The fMRI signal is treated as light emission when rendering a patient-specific high resolution reference MRI volume, obtained at the beginning of the experiment. As a result, the brain glows and emits light from active regions. The low resolution fMRI signal is thus effectively fused with the reference brain with the current transfer function settings yielding an effective focus and context visualization. The delay from a change in the fMRI signal to the visualization is approximately 2 seconds. The advantage of our method over standard 2D slice based methods is shown in a user study. We demonstrate our technique through experiments providing interactive visualization to the fMRI operator and also to the test subject in the scanner through a head mounted display.

We present the first step towards a brain computer interface (BCI) for communication using real-time functional magnetic resonance imaging (fMRI). The subject in the MR scanner sees a virtual keyboard and steers a cursor to select different letters that can be combined to create words. The cursor is moved to the left by activating the left hand, to the right by activating the right hand, down by activating the left toes and up by activating the right toes. To select a letter, the subject simply rests for a number of seconds. We can thus communicate with the subject in the scanner by for example showing questions that the subject can answer. Similar BCI for communication have been made with electroencephalography (EEG). The subject then focuses on a letter while different rows and columns of the virtual keyboard are flashing and the system tries to detect if the correct letter is flashing or not. In our setup we instead classify the brain activity. Our system is neither limited to a communication interface, but can be used for any interface where five degrees of freedom is necessary.

We present a method for fast phase based registration of volume data for medical applications. As the number of different modalities within medical imaging increases, it becomes more and more important with registration that works for a mixture of modalities. For these applications the phase based registration approach has proven to be superior. Today there seem to be two kinds of groups that work with medical image registration, one that works with refining of the registration algorithms and one that works with implementation of more simple algorithms on graphic cards for speeding up the algorithms. We put the work from these groups together and get the best from both worlds. We achieve a speedup of 10-30 compared to our CPU implementation, which makes fast phase based registration possible for large medical volumes.

We present e method for controlling a dynamical system using real-time fMRI. The objective for the subject in the MR scanner is to balance an inverse pendulum by activating the left or right hand or resting. The brain activity is clasified each second by a neural network and the classification is sent to a pendulum simulator to change the state of the pendulum. The state of the inverse pendulum is shown to the subject in a pair of VR goggles. The subject was able to balance the inverse pendulum during a 7 minute test run.

We present a method for controlling a dynamical system using real-time fMRI. The objective for the subject in the MR scanner is to balance an inverted pendulum by activating the left or right hand or resting. The brain activity is classified each second by a neural network and the classification is sent to a pendulum simulator to change the force applied to the pendulum. The state of the inverted pendulum is shown to the subject in a pair of VR goggles. The subject was able to balance the inverted pendulum during several minutes, both with real activity and imagined activity. In each classification 9000 brain voxels were used and the response time for the system to detect a change of activity was on average 2-4 seconds. The developments here have a potential to aid people with communication disabilities, such as locked in people. Another future potential application can be to serve as a tool for stroke and Parkinson patients to be able to train the damaged brain area and get real-time feedback for more efficient training.

The Direct Weight Optimization (DWO) approach is a nonparametric estimation approach that has appeared in recent years within the field of nonlinear system identification. In previous work, all function classes for which DWO has been studied have included only continuous functions. However, in many applications it would be desirable also to be able to handle discontinuous functions. Inspired by the bilateral filter method from image processing, such an extension of the DWO framework is proposed for the smoothing problem. Examples show that the properties of the new approach regarding the handling of discontinuities are similar to the bilateral filter, while at the same time DWO offers a greater flexibility with respect to different function classes handled.

The inverse gradient problem, finding a scalar field f with a gradient near a given vector field g on some bounded and connected domain Omega epsilon R(n), can be solved by means of a Poisson equation with inhomogeneous Neumann boundary conditions. We present an elementary derivation of this partial differential equation and an efficient multigrid-based method to numerically compute the inverse gradient on non-rectangular domains. The utility of the method is demonstrated by a range of important medical applications such as phase unwrapping, pressure computation, inverse deformation fields, and fiber bundle tracking.

Studies of myocardial motion using magnetic resonance imaging usually require multiple breath holds and several methods have been proposed in order to reduce the scan time. Rapid imaging using k-t BLAST has gained much attention with its high reduction factors and image quality. Temporal smoothing, however, may reduce the accuracy when assessing cardiac function. In the present work, a modified reconstruction filter is proposed, that preserves more of the high temporal frequencies. Artificial decimation of a fully sampled data set was used to evaluate the reconstruction filter. Compared to the conventional k-t BLAST reconstruction, the modified filter produced images with sharper temporal delineation of the myocardial walls.  Quantitative analysis by means of regional velocity estimation showed that the modified reconstruction filter produced more accurate velocity estimations.

Tensor based orientation adaptive filtering, an explicit methodology for anisotropic filtering, constitutes a flexible framework for medical image enhancement. The technique features post-filtering steerability and allows user interaction and direct control over the high-frequency contents of the signal. A new class of filters for local structure analysis together with filter networks significantly lowers the complexity to meet the requirements of computation time for clinical use, while maintaining accuracy. In this paper the technique is applied to low-dose CT-images, magnetic resonance angiography and T2-weighted MRI.

Filter networks, i.e. decomposition of a filter set into a layered structure of sparse subfilters has been proven successful for e.g. efficient convolution using finite impulse response filters. The efficiency is due to the significantly reduced number of multiplications and additions per data sample that is required. The computational gain is dependent on the choice of network structure and the graph representation compactly incorporates the network structure in the design objectives. Consequently the graph representation forms a framework for searching the optimal network structure. It also removes the requirement of a layered structure, at the cost of a less compact representation.

This paper presents the new project Efficient Convolution Operators for Image Processing of Volumes and Volume Sequences . The project is carried out in collaboration with Context Vision AB.

By using sequential filtering on 3D and 4D data, the number of nonzero filter coefficients for a desired filter set can be significantly reduced. A sequential convolution structure in combination with a convolver designed for sparse filters is a powerful tool for filtering of multidimensional signals.

The project mainly concerns the design of filter networks, that approximate a desired filter set, while keeping the computational load as low as possible. This is clearly an optimization problem, but it can be formulated in several different ways due to the complexity.

The project is in an initial state and the paper focuses on experiences from prior work and discuss possible approaches for the future progress.

This paper presents a general strategy for automated generation of efficient representations in vision. The approach is highly task oriented and what constitutes the relevant information is defined by a set of examples. The examples are pairs of situations that are dependent through the chosen feature but are otherwise independent. Particularly important concepts in the work are mutual information and canonical correlation. How visual operators and representations can be generated from examples are presented for a number of features, e.g. local orientation, disparity and motion. Interesting similarities to biological vision functions are observed. The results clearly demonstrates the potential of combining advanced filtering techniques and learning strategies based on canonical correlation analysis (CCA).

Digital subtraction angiography, whether based on traditional X-ray or MR, suffers from patient motion artifacts. Until now, the usual remedy is to pixel shift by hand, or in some cases performing a global pixel shift semi-automatically. This is time consuming, and cannot handle rotations or local varying deformations over the image. We have developed a fully automatic algorithm that provides for motion compensation in the presence of large local deformations. Our motion compensation is very accurate for ordinary motions, including large rotations and deformations. It does not matter if the motions are irregular over time. For most images, it takes about a second per image to get adequate accuracy. The method is based on using the phase from filter banks of quadrature filters tuned in different directions and frequencies. Unlike traditional methods for optical flow and correlation, our method is more accurate and less susceptible to disturbing changes in the image, e.g. a moving contrast bolus. The implications for common practice are that radiologists' time can be significantly reduced in ordinary peripheral angiographies and that the number of retakes due to large or local motion artifacts will be much reduced.

This paper presents a general approach for obtaining optimal filters as well as filter sequences. A filter is termed optimal when it minimizes a chosen distance measure with respect to an ideal filter. The method allows specification of the metric via simultaneous weighting functions in multiple domains, e.g. the spatio-temporal space and the Fourier space. It is shown how convolution kernels for efficient spatio-temporal filtering can be implemented in practical situations. The method is based on applying a set of jointly optimized filter kernels in sequence. The optimization of sequential filters is performed using a novel recursive optimization technique. A number of optimization examples are given that demonstrate the role of key parameters such as: number of kernel coefficients, number of filters in sequence, spatio-temporal and Fourier space metrics. In multidimensional filtering applications the method potentially outperforms both standard convolution and FFT based approaches by two-digit numbers.

This paper describes a new algorithm for local orientation estimation. The proposed algorithm detects and separates interfering events in ambiguous neighbourhoods and produces robust estimates of the two most dominant events. A representation suitable for simultaneous representation of two orientations is introduced. The main purpose of this representation is to make averaging of outputs for neigbourhoods containing two orientations possible. The feature extraction is performed by a set of quadrature filters. A method to obtain a large set of quadrature filter responses from a limited basis filter set is introduced. The estimation of the neighbourhood and the separation of the present events are based upon the quadrature responses in terms of local magnitude and phase. The performance of the algorithm is demonstrated using test images.

This thesis concerns robust estimation of low-level features for use in computer vision systems. The presentation consists of two parts.

The first part deals with controllable filters and models. A basis filter set is introduced which supports a computationally efficient synthesis of filters in arbitrary orientations. In contrast to many earlier methods, this approach allows the use of more complex models at an early stage of the processing. A new algorithm for robust estimation of orientation is presented. The algorithm is based on synthesized quadrature responses and supports the simultaneous representation and individual averaging of multiple events. These models are then extended to include estimation and representation of more complex image primitives such as as line ends, T-junctions, crossing lines and curvature. The proposed models are based on symmetry properties in the Fourier domain as well as in the spatial plane and the feature extraction is performed by applying the original basis filters directly on the grey-level image. The basis filters and interpolation scheme are finally generalized to allow synthesis of 3-D filters. The performance of the proposed models and algorithms is demonstrated using test images of both synthetic and real world data.

The second part of the thesis concerns an image feature representation adapted for a robust analogue implementation. A possible use for this approach is in analogue VLSI or corresponding analogue hardware adapted for neural networks. The methods are based on projections of quadrature filter responses and mutual inhibition of magnitude signals.

The multilinear least-squares (MLLS) problem is an extension of the linear least-squares problem. The difference is that a multilinearoperator is used in place of a matrix-vector product. The MLLS istypically a large-scale problem characterized by a large number of local minimizers. It originates, for instance, from the design of filter networks. We present a global search strategy that allows formoving from one local minimizer to a better one. The efficiencyof this strategy isillustrated by results of numerical experiments performed forsome problems related to the design of filter networks.

We present e method for controlling a dynamical system using real-time fMRI. The objective for the subject in the MR scanner is to balance an inverse pendulum by activating the left or right hand or resting. The brain activity is clasified each second by a neural network and the classification is sent to a pendulum simulator to change the state of the pendulum. The state of the inverse pendulum is shown to the subject in a pair of VR goggles. The subject was able to balance the inverse pendulum during a 7 minute test run.

Three-dimensional data processing is becoming more and more common. Typical operations are for example estimation of optical flow in video sequences and orientation estimation in 3-D MR images. This paper proposes an efficient approach to robust low level feature extraction for 3-D image analysis. In contrast to many earlier algorithms the methods proposed in this paper support the use of relatively complex models at the initial processing steps. The aim of this approach is to provide the means to handle complex events at the initial processing steps and to enable reliable estimates in the presence of noise. A limited basis filter set is proposed which forms a basis on the unit sphere and is related to spherical harmonics. From these basis filters, different types of orientation selective filters are synthesized. An interpolation scheme that provides a rotation as well as a translation of the synthesized filter is presented. The purpose is to obtain a robust and invariant feature extraction at a manageable computational cost.