Thomas Vetter

Universität Basel, Bâle, Basel-City, Switzerland

Are you Thomas Vetter?

Claim your profile

Publications (136)146.74 Total impact

  • [Show abstract] [Hide abstract]
    ABSTRACT: We present a generalization of the convolution-based variational image registration approach, in which different regularizers can be implemented by conveniently exchanging the convolution kernel, even if it is nonseparable or nonstationary. Nonseparable kernels pose a challenge because they cannot be efficiently implemented by separate 1D convolutions. We propose to use a low-rank tensor decomposition to efficiently approximate nonseparable convolution. Nonstationary kernels pose an even greater challenge because the convolution kernel depends on, and needs to be evaluated for, every point in the image. We propose to pre-compute the local kernels and efficiently store them in memory using the Tucker tensor decomposition model. In our experiments we use the nonseparable exponential kernel and a nonstationary landmark kernel. The exponential kernel replicates desirable properties of elastic image registration, while the landmark kernel incorporates local prior knowledge about corresponding points in the images. We examine the trade-off between the computational resources needed and the approximation accuracy of the tensor decomposition methods. Furthermore, we obtain very smooth displacement fields even in the presence of large landmark displacements.
    Journal of Mathematical Imaging and Vision 11/2014; 50(3). · 2.33 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Dendritic spines may be tiny in volume, but are of major importance for neuroscience. They are the main receivers for excitatory synaptic connections, and their constant changes in number and in shape reflect the dynamic connectivity of the brain. Two-photon microscopy allows following the fate of individual spines in brain slice preparations and in live animals. The diffraction-limited and non-isotropic resolution of this technique, however, makes detection of such tiny structures rather challenging, especially along the optical axis (z-direction). Here we present a novel spine detection algorithm based on a statistical dendrite intensity model and a corresponding spine probability model. To quantify the fidelity of spine detection, we generated correlative datasets: Following two-photon imaging of live pyramidal cell dendrites, we used serial block-face scanning electron microscopy (SBEM) to reconstruct dendritic ultrastructure in 3D. Statistical models were trained on synthetic fluorescence images generated from SBEM datasets via point spread function (PSF) convolution. After the training period, we tested automatic spine detection on real two-photon datasets and compared the result to ground truth (correlative SBEM data). The performance of our algorithm allowed tracking changes in spine volume automatically over several hours. Using a second fluorescent protein targeted to the endoplasmic reticulum, we could analyze the motion of this organelle inside individual spines. Furthermore, we show that it is possible to distinguish activated spines from non-stimulated neighbors by detection of fluorescently labeled presynaptic vesicle clusters. These examples illustrate how automatic segmentation in 5D (x, y, z, t, λ) allows us to investigate brain dynamics at the level of individual synaptic connections.
    Medical Image Analysis 09/2014; · 3.68 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: A novel stereoscopic image acquisition setup and a procedure for measuring multidimensional particle size distributions (nD PSDs) during crystallization based on image analysis are presented. Images of crystals in suspension passing a flow through cell are generated by two cameras which are arranged in an orthogonal manner. Particles are conveyed to the flow through cell using a sampling loop, thus allowing for online monitoring. Automated image analysis provides contour data which can be used to classify crystals into different generic particle model classes. For each type of particle size data is calculated and stored. Finally, time resolved nD PSD data can be calculated. The accuracy of this novel size measurement was confirmed by comparison to measurements obtained with a Coulter Multisizer. The non-invasive nature and repeatability of experiments are shown by monitoring populations of sodium chloride and of the β polymorph of l-glutamic acid under different conditions. Finally, crystal growth of acetaminophen during cooling crystallization is shown. In addition, a virtual test bench is used to study the measurement method in silico.
    Chemical Engineering Science 02/2014; 105:155–168. · 2.61 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this paper we propose a new approach for spatially-varying registration using Gaussian process priors. The method is based on the idea of spectral tempering, i.e. the spectrum of the Gaussian process is modified depending on a user defined tempering function. The result is a non-stationary Gaussian process, which induces different amount of smoothness in different areas. In contrast to most other schemes for spatially-varying registration, our approach does not require any change in the registration algorithm itself, but only affects the prior model. Thus we can obtain spatially-varying versions of any registration method whose deformation prior can be formulated in terms of a Gaussian process. This includes for example most spline-based models, but also statistical shape or deformation models. We present results for the problem of atlas based skull-registration of cone beam CT images. These datasets are difficult to register as they contain a large amount of noise around the teeth. We show that with our method we can become robust against noise, but still obtain accurate correspondence where the data is clean.
    01/2014; 17(Pt 2):413-20.
  • [Show abstract] [Hide abstract]
    ABSTRACT: A protocol for the estimation of growth kinetics for complex-shaped particles is presented. The estimation is based on multidimensional particle size distribution (nD PSD) and concentration data. While the latter is obtained by an in situ mid-infrared absorption probe, nD PSD data is measured via an imaging based setup presented earlier. The data is fitted to the output of a morphological population balance equation, which is solved by a customized high resolution algorithm. The procedure is first validated in silico using a virtual implementation of the measurement setup before it is applied to seeded desupersaturation experiments of the β polymorph of l-glutamic acid. Prominent broadening of the product PSD is observed and different size (in)dependent growth models are fitted to the data. Confidence intervals, local identifiability, and correlation of the parameters are studied. Finally, the estimated growth rate is compared to literature results.
    Industrial & Engineering Chemistry Research 12/2013; 53(22):9136–9148. · 2.24 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Understanding crystal growth from solution is crucial to control the evolution of crystal morphologies. Experiments, molecular simulations, and theory were combined to examine the morphology of urea crystals grown in different solutions. To get a rational representation of all the possible habits a shape diagram is introduced in which the habit dependence on the relative growth rates is illustrated.
    Angewandte Chemie International Edition 10/2013; 52(50). · 11.34 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this work, we investigate a comprehensive model describing nucleation, growth and Ostwald ripening based on the kinetic rate equation and compare it to commonly used population balance equation models that either describe nucleation and crystal growth or crystal growth and Ostwald ripening. The kinetic rate equation gives a microscopic description of crystallization, i.e., the process is seen as an attachment and detachment of crystals of different sizes to and from each other, thereby changing their size. A hybrid model is employed in which the discrete kinetic rate equation is used to describe the smallest particle sizes while a Fokker–Planck equation is used to approximate the kinetic rate equation at larger particle sizes. This allows us to cover crystals in a size range starting from a single molecule up to macroscopic particle sizes and to solve the model numerically with reasonable computational effort and great accuracy. We show that the model based on the kinetic rate equation describes the processes of nucleation, crystal growth, and Ostwald ripening accurately in a single, continuous model. This is set in contrast with classical population balance equation models that require, due to their underlying assumptions, separation of the process of nucleation from the process of Ostwald ripening. We compare the results of the two models for different sets of parameters (such as different solubilities, surface tensions, initial supersaturations, and seed distributions). Using these results, we assess the advantages and disadvantages of models based on the kinetic rate equation in comparison to models employing a population balance equation.
    Crystal Growth & Design 10/2013; 13(11):4890–4905. · 4.56 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: We present a method to compute the conditional distribution of a statistical shape model given partial data. The result is a "posterior shape model", which is again a statistical shape model of the same form as the original model. This allows its direct use in the variety of algorithms that include prior knowledge about the variability of a class of shapes with a statistical shape model. Posterior shape models then provide a statistically sound yet easy method to integrate partial data into these algorithms. Usually, shape models represent a complete organ, for instance in our experiments the femur bone, modeled by a multivariate normal distribution. But because in many application certain parts of the shape are known a priori, it is of great interest to model the posterior distribution of the whole shape given the known parts. These could be isolated landmark points or larger portions of the shape, like the healthy part of a pathological or damaged organ. However, because for most shape models the dimensionality of the data is much higher than the number of examples, the normal distribution is singular, and the conditional distribution not readily available. In this paper, we present two main contributions: First, we show how the posterior model can be efficiently computed as a statistical shape model in standard form and used in any shape model algorithm. We complement this paper with a freely available implementation of our algorithms. Second, we show that most common approaches put forth in the literature to overcome this are equivalent to probabilistic principal component analysis (PPCA), and Gaussian Process regression. To illustrate the use of posterior shape models, we apply them on two problems from medical image analysis: model-based image segmentation incorporating prior knowledge from landmarks, and the prediction of anatomically correct knee shapes for trochlear dysplasia patients, which constitutes a novel medical application. Our experiments confirm that the use of conditional shape models for image segmentation improves the overall segmentation accuracy and robustness.
    Medical image analysis 06/2013; 17(8):959-973. · 3.09 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: We present a novel method for nonrigid registration of 3D surfaces and images. The method can be used to register surfaces by means of their distance images, or to register medical images directly. It is formulated as a minimization problem of a sum of several terms representing the desired properties of a registration result: smoothness, volume preservation, matching of the surface, its curvature, and possible other feature images, as well as consistency with previous registration results of similar objects, represented by a statistical deformation model. While most of these concepts are already known, we present a coherent continuous formulation of these constraints, including the statistical deformation model. This continuous formulation renders the registration method independent of its discretization. The finite element discretization we present is, while independent of the registration functional, the second main contribution of this paper. The local discontinuous Galerkin method has not previously been used in image registration, and it provides an efficient and general framework to discretize each of the terms of our functional. Computational efficiency and modest memory consumption are achieved thanks to parallelization and locally adaptive mesh refinement. This allows for the first time the use of otherwise prohibitively large 3D statistical deformation models.
    Computational and Mathematical Methods in Medicine 01/2013; 2013:674273. · 0.79 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: We introduce a robust multi-object tracking for abstract multi-dimensional feature vectors. The Condensation and the Wavelet Approximated Reduced Vector Machine (W-RVM) approach are joined to spend only as much as necessary effort for easy to discriminate regions (Condensation) and measurement locations (W-RVM) of the feature space, but most for regions and locations with high statistical likelihood to contain the object of interest. The new 3D Cascaded Condensation Tracking (CCT) yields more than 10 times faster tracking than state-of-art detection methods. We demonstrate HCI applications by high resolution face tracking within a large camera scene with an active dual camera system.
    Computer Standards & Interfaces 11/2012; · 1.18 Impact Factor
  • Thomas Albrecht, Thomas Vetter
    [Show abstract] [Hide abstract]
    ABSTRACT: We present a method to automatically reposition the fragments of a broken bone based on surface meshes segmented from CT scans. The result of this virtual fracture reduction is intended to be used as an operation plan for a medical procedure. Particularly in minimally invasive surgery like intramedullary nailing, the correct repositioning of bone fragments is not always apparent or visible without an operation plan. We propose to achieve automatic fracture reduction by fitting the bone fragments to an intact reference bone mesh with a modified Iterative Closest Point (ICP) algorithm. A suitable reference could be the same patient's contra-lateral bone. In the absence of a CT scan of this bone, we propose to use a statistical shape model as a reference. The shape model is automatically adapted to match the anatomy of the broken bone, apart from the bone's length, which has to be correctly initialized. Our experiments show that we can limit the rotational alignment error to below 5 degrees, compared to 15 degrees in current medical practice.
    Proceedings of the 2012 international conference on Mesh Processing in Medical Image Analysis; 10/2012
  • [Show abstract] [Hide abstract]
    ABSTRACT: Controlling the shape of crystals is of great practical relevance in fields like pharmacology and fine chemistry. Here we examine the paradigmatic case of urea which is known to crystallize from water with a needle-like morphology. To prevent this undesired effect, inhibitors that selectively favor or discourage the growth of specific crystal faces can be used. In urea the most relevant faces are the {001} and the {110} which are known to grow fast and slow, respectively. The relevant growth speed difference between these two crystal faces is responsible for the needle-like structure of crystals grown in water solution. To prevent this effect, additives are used to slow down the growth of one face relative to another, thus controlling the shape of the crystal. We study the growth of fast {001} and slow {110} faces in water solution and the effect of shape controlling inhibitors like biuret. Extensive sampling through molecular dynamics simulations provides a microscopic picture of the growth mechanism and of the role of the additives. We find a continuous growth mechanism on the {001} face, while the slow growing {110} face evolves through a birth and spread process, in which the rate-determining step is the formation on the surface of a two-dimensional crystalline nucleus. On the {001} face, growth inhibitors like biuret compete with urea for the adsorption on surface lattice sites; on the {110} face instead additives cannot interact specifically with surface sites and play a marginal sterical hindrance of the crystal growth. The free energies of adsorption of additives and urea are evaluated with advanced simulation methods (well-tempered metadynamics) allowing a microscopic understanding of the selective effect of additives. Based on this case study, general principles for the understanding of the anisotropic growth of molecular crystals from solutions are laid out. Our work is a step toward a rational development of novel shape-affecting additives.
    Journal of the American Chemical Society 09/2012; 134(41):17221-33. · 11.44 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: A protocol for measuring multidimensional particle size distributions during crystallization is presented. The instrument consists of a flow through cell where it is possible to take images of particles in suspension from two perpendicular directions. The flow through cell is connected to a batch crystallizer through a sampling loop and a dilution system that allow tuning the suspension density in the cell for the sake of image quality. The images thus obtained are analyzed with a fast image analysis algorithm and characteristic lengths of particles are calculated. Generic geometric shapes representing different types of crystals are defined as particle classes. We present a method to classify particles into these classes, which enables for instance to differentiate between different polymorphs of a substance when their shape is different. The capabilities of the measurement device and of the algorithm are illustrated by comparing their performance to standard measurement tools like a Coulter Multisizer. The comparison shows that the two instruments give the same result. The effectiveness and accuracy of the protocol proposed is assessed by monitoring size and shape of crystals of acetylsalicylic acid and of paracetamol during seeded cooling crystallization.
    Chemical Engineering Science 07/2012; 77:130–142. · 2.61 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Using a crossover recognition memory testing paradigm, we tested whether the effects on face recognition of the memorability component of face typicality (Vokey & Read, 1992, 1995) are due primarily to the encoding process occurring during study or to the retrieval process occurring at test. At study, faces were either veridical in form or at moderate (Experiment 1) or extreme (Experiment 2) levels of caricature. The variable of degree of facial caricature at study was crossed with the degree of caricature at test. The primary contribution of increased memorability to increased hit rate was through increased distinctiveness at study. Increased distinctiveness at test contributed to substantial reductions in the false alarm rate, too. Signal detection analyses confirmed that the mirror effects obtained were primarily stimulus/memory-based, rather than decision-based. Contrary to the conclusion of Vokey and Read (1992), we found that increments in face memorability produced increments in face recognition that were due at least as much to enhanced encoding of studied faces as they were to increased rejection of distractor faces.
    Memory & Cognition 04/2012; 28(7):1173-1182. · 1.92 Impact Factor
  • The Insight Journal. 01/2012;
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we introduce concepts to reduce the computational complexity of regression, which are successfully used for Support Vector Machines. To the best of our knowledge, we are the first to publish the use of a cascaded Reduced Set Vector approach for regression. The Wavelet-Approximated Reduced Vector Machine classifiers for face and facial feature point detection are extended to regression for efficient and robust head pose estimation. We use synthetic data, generated by the 3D Morph able Model, for optimal training sets and demonstrate results superior to state-of-the-art techniques. The new Wavelet Reduced Vector Regression shows similarly good results on natural data, gaining a reduction of the complexity by a factor of up to 560. The introduced Evolutionary Regression Tree uses coarse-to-fine loops of strongly reduced regression and classification up to most accurate complex machines. We demonstrate the Cascaded Condensation Tracking for head pose estimation for a large pose range up to +-90 degrees on videostreams.
  • Sixth Workshop on Microscopic Image Analysis with Applications in Biology, Heidelberg; 11/2011
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Previous research has shown high cross-cultural consensus in personality trait judgments based on faces. However, the information that was provided in these studies included extrafacial features, such as hairstyle or clothes. Such styling information can be intentionally chosen by target persons to express who they are. Using a well-developed and validated Western face model, we were able to formalize the static facial information that is used to make certain personality trait judgments, namely, aggressiveness, extroversion, likeability, risk seeking, social skills, and trustworthiness judgments. We manipulated this information in photographs of Asian and Western faces with natural-looking results. Asian and Western participants identified the enhanced salience of all different personality traits in the faces. Asian participants, however, needed more time for this task. Moreover, faces with enhanced salience of aggressiveness, extroversion, social skills, and trustworthiness were better identified by Western than by Asian participants.
    Social Psychological and Personality Science. 11/2011; 2(6).
  • Source
    Thomas Vetter, Marco Mazzotti, Jörg Brozio
    [Show abstract] [Hide abstract]
    ABSTRACT: In the present study, the effect of the polymeric additive Pluronic F127 on the growth rate of Ibuprofen (IBU) crystals from ethanol–water mixtures and its temperature and additive concentration dependence is investigated quantitatively. It is found that the addition of Pluronic F127 (PF127) to the solvent mixture slows the growth rate. At a temperature of 15 °C and an initial supersaturation of 1.5, for instance, 4 and 8 wt % of PF127 lead to a growth rate of 70% and 55%, respectively, of the value without additive. To characterize the growth kinetics, seeded desupersaturation experiments are carried out and monitored using in situ measurement techniques such as attenuated total reflectance Fourier transform infrared (ATR-FTIR) spectroscopy and focused beam reflectance measurement (FBRM). The solubility of IBU at different temperatures and PF127 concentrations is determined by using a multivariate calibration model for the ATR-FTIR signal. A population balance model describing the process is fitted to the experimental data by estimating parameters of the growth rate expression used. The growth rate was identified to be surface integration controlled. It is proposed that the additive causes changes in the growth rate due to isotropic adsorption of the polymer molecules on the crystal surfaces. The purity of the crystals grown in the presence of the additive is evaluated using high performance liquid chromatography (HPLC). It is shown that the PF127 used in the crystallization process is not incorporated into the crystals.
    Crystal Growth & Design. 08/2011; 11(9).
  • Source
    B. Amberg, T. Vetter
    [Show abstract] [Hide abstract]
    ABSTRACT: In video post-production it is often necessary to track interest points in the video. This is called off-line tracking, because the complete video is available to the algorithm and can be contrasted with on-line tracking, where an incoming stream is tracked in real time. Off-line tracking should be accurate and - if used interactively - needs to be fast, preferably faster than real-time. We describe a 50 to 100 frames per second off-line tracking algorithm, which globally maximizes the probability of the track given the complete video. The algorithm is more reliable than previous methods because it explains the complete frames, not only the patches of the final track, making as much use of the data as possible. It achieves efficiency by using a greedy search strategy with deferred cost evaluation, focusing the computational effort on the most promising track candidates while finding the globally optimal track.
    Computer Vision and Pattern Recognition (CVPR), 2011 IEEE Conference on; 07/2011

Publication Stats

6k Citations
146.74 Total Impact Points


  • 2003–2014
    • Universität Basel
      • • Department of Mathematics and Computer Science
      • • Departement Informatik
      Bâle, Basel-City, Switzerland
  • 1994–2012
    • Max Planck Institute for Biological Cybernetics
      Tübingen, Baden-Württemberg, Germany
    • Office of Naval Research
      Arlington, Virginia, United States
  • 2007
    • General Motors Company
      Detroit, Michigan, United States
  • 1994–2003
    • Massachusetts Institute of Technology
      • • Center for Biological and Computational Learning
      • • Department of Brain and Cognitive Sciences
      Cambridge, MA, United States
  • 2002
    • University of Freiburg
      Freiburg, Baden-Württemberg, Germany
  • 1999
    • University of Texas at Dallas
      • School of Behavioral and Brain Sciences
      Richardson, Texas, United States