The SISTA research unit organises monthly seminars, mainly given by SISTA-members or -visitors about their research and recent results. The seminars are scheduled by Lieven De Lathauwer and Johan Suykens. If you want to get more information about these seminars, please contact them.

For on-line directions to get to our department, click here.

Agenda:

Contents

Place:

Session: Industrial Research Projects and Spin-offs 

Chair: Bart De Moor

09.00-09.50 h: Survey of the industrial projects of SISTA, Bart De Moor

In this presentation we explain the different channels through which the research in SISTA gets financed. We will elaborate on what is called 'the third channel', namely that of the so-called direct contract research between the university and an industrial partner.

We will briefly comment on the several industrial projects that SISTA was involved in up to this day, e.g. Glass (Philips), Continuous Variable Transmissions (VCST), Energy and Gaz consumption profiles (Electrabel, Laborelec), Fraud Detection (ASPeCT, KPNO), Modal analysis (SYNOPSIS, FLITE, Flanders Drive, LMS), Space Module Control (PROBA, Verhaert), Bio informatics (STWW-VIB), Steel Oven Control (CRM), Cable modems (Siemens-ATEA), ADSL/VDSL modems (Alcatel-Bell & Alcatel-Microelectronics),acoustic echo and noise reduction systems (Lernout & Hauspie and Philips-ITCL), PDT Coil. We will also briefly elaborate on the policy of this university with respect to valorisation of research, by selling licences, patenting inventions or creating spin-off companies. Two of our own spin-off companies will present themselves. 

10.00-10.50 h: Peter Van Overschee, CEO  ISMC NV 

ISMC NV, Technologielaan 11/0101, B-3000 Leuven Belgium

T: +32-016393087

F: +32-016393080 

E: peter.vanoverschee@ismc.be

W: www.ismc.be 

ISMC is an engineering company with specialists in the fields of process identification, information technology, model predictive control, real time information, processing, and integration of simulation and control technology. ISMC was founded in April 1995 as a spin-off of the Electrical Engineering department at the Katholieke Universiteit Leuven (Belgium). The expertise of ISMC resulted in advanced modeling and control products and in a multitude of consulting, control and information systems projects realized at production sites of chemical processing industries, power generating industries and oil exploration companies. 

In this seminar we will introduce the history of ISMC and illustrate the road from an initial engineering based spin-off company to a mature product driven company. We will give an overview of the current ISMC products and the technology used in these products. We will also indicate how university based research contributes (past/present and future) to the development of the ISMC products.

11.00-11.50 h: Herman Verrelst, CEO Data4s NV

Data4s NV, Kardinaal Mercierlaan 94

B-3001 Leuven Belgium

T: +32-016321801

F: +32-016321970

E: herman.verrelst@data4s.com

W: www.data4s.com ; www.sport4s.com 

Data4S N.V. is a high-tech SME incorporated January 2000 as a spin-off company from the Katholieke Universiteit of Leuven, more specific the SISTA/COSIC laboratory of the Departement of Electrical Engineering. Its aim is to commercialize the research results of the group in the field of artificial intelligence, data mining and profiling algorithms. From a common technological and knowledge platform, Data4s is first of all active in the market of Customer Relation Management (CRM) in telecommunications, Internet, banking, etc. The main references of Data4s competence and success are the collaboration with major industrial companies like world leading TelCo company Vodafone, Belgian's third GSM network operator, KPN Orange and Europe's leading Payment Service Provider, Europay International. 

The second major activity is in the field of bio-informatics: Data4s bioinformatics' mission is to be a world leader in diagnostics and disease management software that integrates expert medical knowledge with genetic sequence and expression data (as obtained from current and future high throughput genomics devices). In this field, Data4s holds a close collaboration with the MicroArray Facility (MAF) of the Flemish Institute of Biotechnology (VIB) and with the University Hospitals of Leuven. 

In the seminar we would like to present the technology and know-how underlying Data4s' activities. We will give some details on the incorporation process of Data4s beginning this year and the achievements in our first year of commercial activity. 

Place:KULeuven ESAT 00.62

Session: Bioinformatics 

Chair:  Dr. Ir. Yves Moreau

09.00-09.20h: Bioinformatics at ESAT-SISTA - Yves Moreau (K.U.Leuven ESAT-SISTA)

We give a brief overview of the different activities in the field of bioinformatics at SISTA.

09.20-10.00h: Finding regulatory motifs in DNA sequences by Gibbs sampling -

Gert Thijs (K.U.Leuven ESAT-SISTA)

Transcriptome analysis allows to detect and cluster genes that are coexpressed under various biological circumstances. Using this information, it is possible to investigate the cis-acting sequences controlling the transcription of these genes. We extended the Gibbs Sampling algorithm for motif finding with a higher-order background model to improve the robustness of the algorithm to noisy data. For Arabidopsis thaliana a background model based on a set of carefully selected intergenic sequences was constructed. The use of such higher-order model considerably enhances the performance of our motif finding algorithm in the presence of noisy data. Data sets, in which the regulatory elements are known, were used to test the influence of the different background models on the performance of the motif detection algorithm. To demonstrate the performance on a real life problem, we analyzed a microarray data set coming from physical wounding experiments in Arabidopsis thaliana.

10.00-10.10h: break

10.10-10.50h: Adaptive quality-based clustering of gene expression profiles -Frank De Smet (K.U.Leuven ESAT-SISTA)

Clustering genes based on their expression behaviour/profiles is an important step preceding further analysis of the interaction between these genes. Based on the hypothesis that similarity in expression (coexpression) implies similarity in regulatory mechanisms (coregulation), cluster algorithms only have to group genes with a significant degree of coexpression. Other genes have to be excluded from further analysis. With these remarks in mind we propose an iterative two-step algorithm:

First, we try to find an area in the data where the 'density' of expression profiles is locally maximal (based on a preliminary estimate of the radius of the cluster - quality based approach). In a second step, we derive the true radius (or quality) of the cluster. This is achieved by fitting a model to the data using an EM-algorithm. The model used assumes that the data is normalized (this should always be the case when using gene expression profiles). By inferring the radius or quality from the data itself, the biologist is released from estimating this parameter manually (this parameter was sometimes hard to predict).

The most important properties of this approach are:

1. The number of user-defined parameters is minimal (e.g., the number of clusters does not have to be known in advance).

2. Not all genes are assigned to a cluster.

10.50-11.00h: break

11.00-12.00h: Use/need of BioInformatics in a MicroArray Facility -Paul Van Hummelen (VIB MicroArray Facility), Patrick Glenisson (K.U.Leuven ESAT-SISTA)

With the human genome project reaching its completion scientists are looking for new ways to process the sequence data to functional knowledge of the genes in biological processes. Recently a new technique, MicroArrays/DNA-Chip, emerged that allows massive and parallel analysis of gene-expression. One MicroArray typically contains information of ~5,000 genes and an average experiment uses 5-10 arrays not including replicates. It is clear that MicroArrays generate several magnitudes more data than biologists are use to. We estimate that at VIB MicroArray facility 2.4 gigabyte of data will be generated from 300 experiments per year. Not only processing and mining that large amounts of data need new bioinformatics tools, but also for the production process of the MicroArrays itself bioinformatics is indispensable.

In this presentation an overview will be given of applications and developments of informatics in the Laboratory Information Managing System (LIMS) of the MicroArray production process. In addition, image analysis software and data visualization software for large amounts of data points will be presented through biological experiments performed at the facility.

Finally, the issues of data analysis and remote data mining will be discussed.

Place: Room 02.58, Department Electrical Engineering (ESAT), K.U.Leuven

Session: Pattern recognition of brain tumours based on magnetic resonance spectra. Rosemary Tate from thge dept. of computer science of Sussex and ST George Hospital Medical School, London, U.K.

Chair: prof. Sabine Van Huffel

Time: 10h30-11h15

Place: Room 02.58, Department Electrical Engineering (ESAT), K.U.Leuven

Session:  Quantum Information Theory

Chair: Dr. Ir. Koen Audenaert

9h00 - 9h50: ``The local manipulation of entanglement: An application of majorization and matrix analysis in quantum physics'' Martin Plenio (Imperial College, London, UK):

In this talk I would like to introduce you to the exciting study of entanglement in quantum physics. Entanglement is a strange form of correlations between distant quantum systems which has recently become to be regarded as a physical resource that can be consumed to allow for the implementation of interesting quantum tasks such as quantum cryptography, quantum teleportation, the construction of ultrasensitive atomic clocks or optical resolution below the diffraction limit. As any other resource, you can convert entanglement from one form to another, you may purify or dilute it and like with any other resource it is important to learn what can be achieved and what cannot be achieved. As we are generally interested in tasks that involve distant parties we would like to see how entanglement can be manipulated when the two parties act individually and communicate classically (eg a telephone). This constraint makes the problem sensible, interesting and pretty tricky. It is becoming more and more clear that, fortunately, the methods of majorization and matrix analysis provide many useful tools for this study.

I hope that with this talk I can convince you that this is an interesting field of study and maybe I can convince some of you to join us in the quest to gain a full theory of quantum entanglement.

9h50-10h00: break

10h00-10h50:

``Thompson's theorems, Majorisation and the Calculation of Entanglement'' Koenraad Audenaert (K.U. Leuven, ESAT-SISTA)

Entanglement is a characteristic of quantum systems that has received much attention recently (cfr. Martin Plenio's talk). Several measures quantifying the amount of entanglement have been proposed, but most of them are hard to compute. A particularly important one is the so-called entanglement of formation (EoF). For the simplest systems (two spin-1/2 particles), an analytic expression for the EoF is known (the "Wootters formula"). Attempts at generalising this formula to higher-dimensional systems have so far remained futile.

In this talk we will first give a short proof of this formula by showing its connection with a celebrated theorem by R.C. Thompson. This theorem gives the necessary and sufficient conditions for the existence of a symmetric matrix with prescribed singular values and diagonal elements.

The main part of the talk will concentrate on this and other theorems of Thompson. The connection with majorisation and group theory will be highlighted and we will discuss possible avenues for investigating and, hopefully, generalising these theorems.

10h50-11h00: break

11h00-11h50: ``Local filtering operations on 2 qubits'' Frank Verstraete (K.U. Leuven, ESAT-SISTA)

 

Place: 00.62, Electrical Engineering Departement, Kasteelpark Arenberg 10, Leuven

Session: Signal Processing

Chair: Dr. Ir. Geert Leus

09.00-09.50h: Multiuser Communications and Application to Powerlines Thierry Sartenaer (UCL, TELE)

09.50-10.00h: break

10.00-10.25h: Constraints in Channel Shortening Equalizer Design for DMT-Based Systems Geert Ysebaert (K.U.Leuven, ESAT/SISTA)

10.25-10.50h: On the Relationship Between Frequency Domain System Identification and the Equalization of a DMT/OFDM System Koen Vanbleu (K.U.Leuven, ESAT/SISTA)

10.50-11.00h: break

11.00-11.25h: On Wireless Personal Area Networks, a Tentative Architecture Luc Deneire (IMEC/DESICS)

11.25-11.50h: A Reduced Complexity Deterministic Blind Transceiver with Space-Only Block Coding in a Multi-User MIMO Context with Severe Multipath. Olivier Rousseaux (K.U.Leuven, ESAT/SISTA)

Place:

Session: Biomedical Data Processing

Chair: Prof. Dr. Ir. Sabine Van Huffel

09.00-09.50h: An Overview of BIOMEDICAL DATA PROCESSING applications in SISTA, Sabine Van Huffel (K.U.Leuven, ESAT-SISTA)

The following biomedical applications and underlying computational problems, making use of linear algebra, signal processing and system identification, will be introduced in this lecture: 

    - Quantitation of the kidney impulse response in renography,

    - Quantitation of metabolite concentrations and images using in-vivo

Magnetic Resonance Spectroscopic data,

    - Quantitation of brain oxygenation using Near-Infrared Spectroscopy,

    - Quantitation of tactile sensation through oral implants using

Trigeminal Somatosensory Evoked Potentials,

    - Preoperative classification of ovarian tumours, and

    - Fetal Electro-CardioGram Extraction from abdominal recordings.

09.50-10.00h: break

10.00-10.50h: Nonlinear analysis of cardiac rhythm towards clinical

applications, Pierre-Antoine Absil (Montefiore Institute, University of Liege)

This presentation is dedicated to the study of cardiac rhythm. The principal application considered is the prognosis of the future health of patients suffering from congestive heart failure, through the study of long range correlations and scale invariance in their cardiac rhythm.

10.50-11.00h: break

11.00-11.25h: A unified comparison of filtering approaches to water peak suppression in MRS, Leen Vanhamme (K.U. Leuven, ESAT-SISTA)

In biomedical applications of proton MRS, the large water resonance has to be suppressed to allow an adequate quantitation of th metabolite signals. As one usually intends to quantify the latter, suppressing the water peak should not affect them. Most of the water removal procedures that take the measured MRS signal as an input, rely implicity or explicitly on the signal processing concept of filtering. However, several filter proposals lack a theoretical perspective and havenot been explicitly applied to the quantitation of FID signals. We present a unified framework for the comparison of wavelet/Gabor transform based methods, the method of Marion et al., the method of Sodano and Delepierre and the FIR filter based method developed by Sundin et al.. Furthermore, we compare the final parameter estimates obtained by quantifying the output of each water removal method. Although all compared methods are based on the concept of filtering, this study showed clear differences in obtained results. 

From a signal processing perspective, the methods mainly differ in the treatment of the first FID points. Futhermore this study emphasizes two key points. First, FID quantitation must take into account the spectral line removal process if one intends to estimate the metabolite concentrations without bias and with high accuracy. Second, among the four filtering techniques, the most efficient and accurate one in terms of quantitation after spectral line removal, is the one proposed by Sundin et al.. It is the only one that is fully automatic, it is unbiased and performs very close to what is theoretically possible.

11.25-11.50h: Quantification of NAA, Creatine and Choline in Proton MR Turbo Spectroscopic Imaging using Time Domain Fitting, Rene in 't Zandt (K.U. Leuven, ESAT-SISTA)

The use of fast spin echo sequences for spectroscopic imaging (SI) results in a reduced spectroscopic resolution and truncation artifacts due to the limited time available to acquire the full echo signal. Especially at echo time spacings of 136 ms and smaller, clinically relevant markers like creatine and choline become indistinguishable and may hamper the analysis of these resonances. The aim of our study was to demonstrate the feasibility of fast and accurate spectroscopic imaging, by combining Turbo spin echo SI acquisition (TSI) and time-domain processing of the MR signals. All signals were processed assuming a Lorentzian lineshape model function. Despite the limited TSI spectral resolution for 128 points acquisition of a full echo, a time domain fitting analysis provides choline/creatine (0.76 ± 0.04) and NAA/creatine (1.27 ± 0.03) ratios, which are not significantly different from the high spectroscopic resolution in single voxel experiments (0.70 ± 0.07 and 1.25 ± 0.05 respectively).

Place: 01.60, Electrical Engineering Departement, Kasteelpark Arenberg 10, Leuven

Chair: prof.dr.ir. Marc Moonen 

14.30-... : Efficient Least-Squares Adaptive Filters for non-Tapped Delay-Line Structures,  Prof. Ali H. Sayed, Electrical Engineering Dept. UCLA (www.ee.ucla.edu/asl)

This talk provides an overview of recent work on efficient adaptive filters of the least-squares type for filter structures that are not restricted to tapped-delay lines. While it is commonly believed in the literature that shift-structure in the data is necessary for the derivation of fast least-squares algorithms, this work shows that fixed-order and order-recursive filters can still be derived for more general data structures. The results are applied in particular to Laguerre filters.

Place: 00.62, Electrical Engineering Departement, Kasteelpark Arenberg 10, Leuven

Session: Signal Processing

Chair: prof.dr.ir. Sabine Van Huffel

 11h-11h30, Introduction to the Signal Processing Systems group at the TU Eindhoven, Prof. dr. Ir. Jan Bergmans, TU Eindhoven

The Signal Processing Systems group at Eindhoven University of Technology (TU/e) forms part of the department of electrical engineering, and also participates actively in the department of biomedicalengineering. This presentation provides an introduction to the group, with emphasis on its biomedical activities.

The presentation will cover the structure of the group, its role in both departments, its educational activities, its current research programme, and emerging research plans.

11h30-12h: Neuromonitoring research at the Signal Processing Systems group at the TU Eindhoven,  Prof. Pierre Cluitmans, TU Eindhoven

The neuromonitoring research within the SPS group of Tu/e aims at improving methods for the assessment of the state of the central nervous system in patients during anaesthesia and intensive care. In this presentation, an overview of past and present work in this area will be given.

The original research started in the mid 1980s and was focused around the area of monitoring depth of anaesthesia using auditory evoked potentials (AEP). A high-frequency stimulation technique using trains of clicks having random inter-stimulus intervals was developed and evaluated. The method allows assessment of non-linear properties of the auditory pathway and it was shown that this technique allows extracting an average AEP waveform with an acceptable SNR in about 30% of the data acquisition time necessary to obtain the same SNR in the classical AEP method.

Successive research projects focused on establishing a relation between changes in AEP and EEG derived parameters on one hand and depth of anaesthesia/level of sedation on the other, automated pattern recognition in AEP waveforms using artificial neural networks and automated signal validation in clinical EEG signals. Since the early 1990’s a close collaboration with Tilburg University exists in which the neuro-cognitive basis of selective attention and motor preparation is investigated using event related potential paradigms.

Current work on estimation and classification problems in neuromonitoring issues involve pattern recognition and classification methods, single trial event related potential methods by integration of neurophysiological knowledge and spatio-temporal averaging. In addition, other adaptive signal processing methods that we expect to allow an improvement of the current state of the art in neuromonitoring are developed.

Place:00.62, Electrical Engineering Departement, Kasteelpark Arenberg 10, Leuven

Session: Modeling and Controle

Chair: Prof. Dr. Ir. Johan Suykens

09.00-09.25h: Gene Interaction Modelling from Expression Data and Medline Abstracts: Learning Bayesian Networks from Experimental Data and Scientific Literature, Geert Fannes (K.U. Leuven, ESAT-SISTA)

09.25-09.50h: Modeling the acoustic guitar: parameter estimation for dual polarization and coupling, Axel Nackaerts (K.U. Leuven, ESAT-SISTA)

09.50-10.00h: break

10.00-10.25h: Misfit versus latency, Philippe Lemmerling (K.U. Leuven, ESAT-SISTA)

10.25-10.50h: Model predictive control and convex optimization, Jeroen Buijs (K.U. Leuven, ESAT-SISTA)

10.50-11.00h: break

11.00-11.25h: Nonlinear modelling and Automatic Relevance Determination, Maarten Duhoux (K.U. Leuven, ESAT-SISTA)

11.25-11.50h: Standardization of MicroArray Data Exchange. Patrick Glenisson (K.U. Leuven, ESAT-SISTA)

Place:00.62, Electrical Engineering Departement, Kasteelpark Arenberg 10, Leuven

Session: Newcomers Progress I

Chair: Prof. Dr. Ir. Johan Suykens

09.00-09.25h: Mathematical poles in linear models and enhanced state estimation for bilinear systems -Ivan Goethals (K.U.Leuven ESAT-SISTA)

09.25-09.50h: Robust and adaptive design for linearly parameterized systems -Ivan Markovsky (K.U.Leuven ESAT-SISTA)

09.50-10.00h: break

10.00-10.25h: Complex-valued neurons -Luc Hoegaerts (K.U.Leuven ESAT-SISTA)

10.25-10.50h: LS-SVMlab: a Performant Matlab Toolbox for Least Squares Support Vector Machines -Kristiaan Pelckmans (K.U.Leuven ESAT-SISTA)

10.50-11.00h: break

11.00-11.25h: Automatic Processing of Data from Prostate Magnetic Resonance Spectroscopy -Pieter Pels (K.U.Leuven ESAT-SISTA)

11.25-11.50h: Quantification and Classification of MR spectra from human brain tumors using metabolite concentrations as inputs -Andy Devos (K.U.Leuven ESAT-SISTA)

Place:00.62, Electrical Engineering Departement, Kasteelpark Arenberg 10, Leuven

Session: Newcomers Progress II

Chair: Prof. Dr. Ir. Johan Suykens

Send mail to sistawww@esat.kuleuven.ac.be with questions or comments about this web site.
Copyright © 1998 SISTA
Last modified: 17-04-2001