Sigurdur Sigurdsson
Technical University of Denmark
Network
Latest external collaboration on country level. Dive into details by clicking on the dots.
Publication
Featured researches published by Sigurdur Sigurdsson.
IEEE Transactions on Biomedical Engineering | 2004
Sigurdur Sigurdsson; Peter Alshede Philipsen; Lars Kai Hansen; Jan Larsen; Monika Gniadecka; Hans Christian Wulf
Skin lesion classification based on in vitro Raman spectroscopy is approached using a nonlinear neural network classifier. The classification framework is probabilistic and highly automated. The framework includes a feature extraction for Raman spectra and a fully adaptive and robust feedforward neural network classifier. Moreover, classification rules learned by the neural network may be extracted and evaluated for reproducibility, making it possible to explain the class assignment. The classification performance for the present data set, involving 222 cases and five lesion types, was 80.5%/spl plusmn/5.3% correct classification of malignant melanoma, which is similar to that of trained dermatologists based on visual inspection. The skin cancer basal cell carcinoma has a classification rate of 95.8%/spl plusmn/2.7%, which is excellent. The overall classification rate of skin lesions is 94.8%/spl plusmn/3.0%. Spectral regions, which are important for network classification, are demonstrated to reproduce. Small distinctive bands in the spectrum, corresponding to specific lipids and proteins, are shown to hold the discriminating information which the network uses to diagnose skin lesions.
ICA | 2000
Thomas Kolenda; Lars Kai Hansen; Sigurdur Sigurdsson
Automatic content-based classification of text documents is highly important for information filtering, searching, and hyperscripting. State-of-the-art text mining tools are based on statistical pattern recognition working from relatively basic document features such as term frequency histograms. Since term lists are high-dimensional and we typically have access to rather limited labeled databases, representation becomes an important issue. The problem of high dimensions has been approached with principal component analysis (PCA) — in text mining called latent semantic indexing (LSI) [4]. In this chapter we will argue that PCA should be replaced by the closely related independent component analysis (ICA). We will apply the ICA algorithm presented in Chapter 9 which is able to identify a generalizable low-dimensional basis set in the face of high-dimensional noisy data. The major benefit of using ICA is that the representation is better aligned with the content group structure than PCA. We apply our ICA technology to two public domain data sets: a subset of the MED medical abstracts database and the CRAN set of aerodynamics abstracts. In the first set we find that the unsupervised classification based on the ICA conforms well with the associated labels, while in the second set we find that the independent text components are stable but show less agreement with the given labels.
Journal of Applied Physiology | 2008
Svend Strandgaard; Sigurdur Sigurdsson
It is intuitively clear that the cerebral circulation cannot take part in general cardiovascular regulation. During states of shock, where the sympathetic nervous system is activated, leading to a decrease in the perfusion of kidneys and mesenteric vascular bed, the cerebral circulation is kept
international conference on acoustics, speech, and signal processing | 2000
Lars Kai Hansen; Sigurdur Sigurdsson; Thomas Kolenda; Finn Årup Nielsen; Ulrik Kjems; Jan Larsen
We apply and discuss generalizable Gaussian mixture (GGM) models for text mining. The model automatically adapts model complexity for a given text representation. We show that the generalizability of these models depends on the dimensionality of the representation and the sample size. We discuss the relation between supervised and unsupervised learning in the test data. Finally, we implement a novelty detector based on the density model.
international conference on acoustics, speech, and signal processing | 2002
Sigurdur Sigurdsson; Jan Larsen; Lars Kai Hansen; Peter Alshede Philipsen; Hans Christian Wulf
We extend MacKays Bayesian approach to neural classifiers to include an outlier detector mechanism. We show that the outlier detector can locate misclassified samples.
international conference on acoustics, speech, and signal processing | 2007
Andreas Brinch Nielsen; Sigurdur Sigurdsson; Lars Kai Hansen; Jerónimo Arenas-García
Automatic knowledge extraction from music signals is a key component for most music organization and music information retrieval systems. In this paper, we consider the problem of instrument modelling and instrument classification from the rough audio data. Existing systems for automatic instrument classification operate normally on a relatively large number of features, from which those related to the spectrum of the audio signal are particularly relevant. In this paper, we confront two different models about the spectral characterization of musical instruments. The first assumes a constant envelope of the spectrum (i.e., independent from the pitch), whereas the second assumes a constant relation among the amplitude of the harmonics. The first model is related to the Mel frequency cepstrum coefficients (MFCCs), while the second leads to what we will refer to as harmonic representation (HR). Experiments on a large database of real instrument recordings show that the first model offers a more satisfactory characterization, and therefore MFCCs should be preferred to HR for instrument modelling/classification.
Journal of Cerebral Blood Flow and Metabolism | 2014
Sigurdur Sigurdsson; Olaf B. Paulson; Arne Høj Nielsen; Svend Strandgaard
The lower limit of autoregulation of cerebral blood flow (CBF) can be modulated with both angiotensin-converting enzyme (ACE) inhibitors and angiotensin receptor blockers (ARB). The influence of bradykinin antagonism on ARB-induced changes was the subject of this study. CBF was measured in Sprague–Dawley rats with laser Doppler technique. The blood pressure was lowered by controlled bleeding. Six groups of rats were studied: a control group and five groups given drugs intravenously: an ACE inhibitor (enalaprilat), an ARB (candesartan), a bradykinin-2 receptor antagonist (Hoe 140), a combination of enalaprilat and Hoe 140, and a combination of candesartan and Hoe 140. In the control group, the lower limit of CBF autoregulation was 54±9 mm Hg (mean±s.d.), with enalaprilat it was 46±6, with candesartan 39±8, with Hoe 140 53±6, with enalaprilat/Hoe 140 52±6, and with candesartan/Hoe 140 50±7. Both enalaprilat and candesartan lowered the lower limit of autoregulation of CBF significantly. The bradykinin antagonist abolished not only the effect of the ACE inhibitor but surprisingly also the effect of the ARB on the lower limit of CBF autoregulation, the latter suggesting an effect on intravascular bradykinin.
Journal of Applied Physiology | 2008
Svend Strandgaard; Sigurdur Sigurdsson
to the editor: This has been an interesting debate that has narrowed the field of disagreement between the two camps. Despite our title, we described in our first article how the sympathetic perivascular nerves, although not influencing steady-state autoregulation at rest, can be activated to
Neural Networks for Signal Processing X. Proceedings of the 2000 IEEE Signal Processing Society Workshop (Cat. No.00TH8501) | 2000
Sigurdur Sigurdsson; Jan Larsen; Lars Kai Hansen
Modeling with flexible models, such as neural networks, requires careful control of the model complexity and generalization ability of the resulting model which finds expression in the ubiquitous bias-variance dilemma. Regularization is a tool for optimizing the model structure reducing variance at the expense of introducing extra bias. The overall objective of adaptive regularization is to tune the amount of regularization ensuring minimal generalization error. Regularization is a supplement to direct model selection techniques like step-wise selection and one would prefer a hybrid scheme; however, a very flexible regularization may substitute the need for selection procedures. This paper investigates recently suggested adaptive regularization schemes. Some methods focus directly on minimizing an estimate of the generalization error (either algebraic or empirical), whereas others start from different criteria, e.g., the Bayesian evidence. The evidence expresses basically the probability of the model, which is conceptually different from generalization error; however, asymptotically for large training data sets they will converge. First the basic model definition, training and generalization is presented. Next, different adaptive regularization schemes are reviewed and extended. Finally, the experimental section presents a comparative study concerning linear models for regression/time series problems.
Stroke | 2017
Sigurdur Sigurdsson; Thor Aspelund; Olafur Kjartansson; Elias F. Gudmundsson; Maria K. Jonsdottir; Gudny Eiriksdottir; Palmi V. Jonsson; Mark A. van Buchem; Vilmundur Gudnason; Lenore J. Launer
Background and Purpose— The differentiation of brain infarcts by region is important because their cause and clinical implications may differ. Information on the incidence of these lesions and association with cognition and dementia from longitudinal population studies is scarce. We investigated the incidence of infarcts in cortical, subcortical, cerebellar, and overall brain regions and how prevalent and incident infarcts associate with cognitive change and incident dementia. Methods— Participants (n=2612, 41% men, mean age 74.6±4.8) underwent brain magnetic resonance imaging for the assessment of infarcts and cognitive testing at baseline and on average 5.2 years later. Incident dementia was assessed according to the international guidelines. Results— Twenty-one percent of the study participants developed new infarcts. The risk of incident infarcts in men was higher than the risk in women (1.8; 95% confidence interval, 1.5–2.3). Persons with both incident and prevalent infarcts showed steeper cognitive decline and had almost double relative risk of incident dementia (1.7; 95% confidence interval, 1.3–2.2) compared with those without infarcts. Persons with new subcortical infarcts had the highest risk of incident dementia compared with those without infarcts (2.6; 95% confidence interval, 1.9–3.4). Conclusions— Men are at greater risk of developing incident brain infarcts than women. Persons with incident brain infarcts decline faster in cognition and have an increased risk of dementia compared with those free of infarcts. Incident subcortical infarcts contribute more than cortical and cerebellar infarcts to incident dementia which may indicate that infarcts of small vessel disease origin contribute more to the development of dementia than infarcts of embolic origin in larger vessels.