Programme

14h00-14h30: Arnaud Installé, ESAT-SCD (SISTA)
14h30-14h40: Questions and Discussion

14h40-15h10: Ben Van Calster, Department of Development & Regeneration, KU Leuven 15h10-15h20: Questions and Discussion

15h20-15h40: Coffee Break

15h40-16h10: Kirsten Van Hoorde, ESAT-SCD (SISTA)
16h10-16h20: Questions and Discussion

16h20-16h50: Laure Wynants, ESAT-SCD (SISTA)
16h50-17h00: Questions and Discussion

Perinatology is the medical field that is concerned with the health of mother and child before, during and after birth. Recent decades have seen important advances in perinatal monitoring technology, aimed e.g. at early detection of fetal asphyxia and impending premature birth, assessing fetal condition during delivery, assessing neonatal cerebral autoregulation status, and improving neonatal comfort. This symposium provides a survey of the state of the art and trends in this field, both from a technological and from a clinical perspective. The symposium serves also to honor Prof. Sabine Van Huffel, one of the leading specialists in perinatal monitoring, on the occasion of the award of her honorary doctorate by Eindhoven University of Technology.

Surrounding the symposium there will be an exhibition with scientific posters and real-life demonstrations (e.g. a smart neonatal mattress and jacket, serious obstetric game, electrophysiological pregnancy monitoring).

Jury

• Prof. dr. ir. Yves Willems, chairman
• Prof. dr. ir. S. Van Huffel, promotor
• Prof. dr. Bart Nuttin, promotor
• Prof. dr. ir. Carmen Bartic, promotor
• Prof. dr. ir. Mirjana Popovic (Belgrade University, Serbia)
• Prof. dr. ir. Maarten De Vos (University of Oldenburg, Germany)
• Dr. ir. Joleen H. Blok (Department of Clinical Neurophysiology, Erasmus MC, The Netherlands)
• Prof. dr. ir. Marc Van Hulle
• Prof. dr. ir. Robert Puers

BIOMED workshop

Challenges in Biomonitoring

Thursday, February 21 2013, from 14h till 17h

PLACE:

KU Leuven, Department of Electrical Engineering, Kasteelpark Arenberg 10, 3001 Leuven,  room 01.60

Programme:

14h00-14h30:  Prof. Qun Wan, Chengdu, China

14h30-14h50: Yipeng Liu, ESAT-SCD

14h50-15h00: Questions and Discussion

15h00-15h20:  Devy Widjaja, ESAT-SCD

15h20-15h40:  Tim Willemen, BMe and ESAT-SCD

15h40-15h50: Questions and Discussion

15h50-16h10: Coffee Break

16h10-16h30: Carolina Varon, ESAT-SCD

16h30-16h50: Milica Milosevic, ESAT-SCD

16h50-17h00: Questions and Discussion

Program

9.45 - 10.00: Introduction
10.00 - 10.30: Prof. B. Ceulemans (UZ Antwerpen): Where it all started. The clinical view!
10.30 - 11.00: Prof. H. Sørensen (Technical University of Denmark ): Multi-modal detection of motor seizures
11.00 - 11.30: Prof. H. Aghajan (Stanford University ): Smart Environments and Ambient Intelligence
11.30 - 12.00: Discussion
12.00 - 13.00: Sandwich lunch
13.00 - 15.00: Kris Cuppens: Detection of Epileptic Seizures based on Videoand Accelerometer Recordings -PhD defense-
15.00: Reception

Programme

14h00-14h30: Sofie Van Cauter
14h30-15h00: Anca Croitor
15h00-15h30: Alexander Caicedo
15h30-15h45: Questions and Discussion

15h45-16h015: Break

16h15-16h45: Borbala Hunyadi
16h45-17h15: Kris Cuppens
17h15-17h30: Questions and Discussion

Dr. Jan Luts, School of Mathematical Sciences, University of Technology Sydney

Abstract:

Semiparametric regression is an extension of regression that permits incorporation of flexible functional relationships using basis functions, such as splines and wavelets, and penalties and is now well-developed for cross-sectional, longitudinal and spatial data. Almost all semiparametric regression analyses process the data in batch. That is, a data set is fed into a semiparametric regression procedure at some point in time after its collection. This talk discusses doing semiparametric regression in real time, with data processed as it is collected and made immediately available via modern telecommunications technologies. Regression summaries may be thought of as dynamic web-pages or iDevice apps rather than static tables and figures on a piece of paper. Online processing of data is an old idea and has a very large literature. Our work uses Bayesian approaches, that handle automatic choices of smoothing parameters, and make use of fast variational approximations that are amenable to online updating. This talk represents joint research with Professor Matt P. Wand.

"Second-Order Multidimensional ICA: Theory and Methods"

Dana Lahat (Tel-Aviv University)

Abstract:

Independent component analysis (ICA) and blind source separation (BSS) deal with extracting a number of mutually independent elements from a set of observed linear mixtures. Motivated by various applications, this talk considers a more general and more flexible model: the sources can be partitioned into groups exhibiting dependence within a given group but independence between two different groups. We argue that this is tantamount to considering multidimensional components, as opposed to the standard ICA case which is restricted to one-dimensional components.

Multidimensional data may occur due to various complex relations within the dependent elements. The dimension of a dependent group may not always reflect the actual number of its underlying elements. As a result, in multidimensional models, there is not always a physically meaningful interpretation to separating the multidimensional components back into single-dimensional elements.

The core of this work is the statistical analysis of the blind separation of multidimensional components based on second-order statistics, in a piecewise-stationary model. We develop the likelihood and the associated estimating equations for the Gaussian case. We obtain closed-form expressions for the Fisher information matrix and the Cramér-Rao lower bound (CRLB) of the de-mixing parameters, as well as the mean square error (MSE) of the component estimates. For Gaussian data, our separation criterion achieves, up to higher-order terms, the CRLB, and is thus optimal in the MSE sense.

We present necessary and sufficient conditions for the model to be identifiable. These are also the sufficient and necessary conditions for joint block diagonalization (JBD) of any set of real positive-definite symmetric matrices to be unique.

We then turn to the case when the separation procedure is based on a one-dimensional model, followed by a clustering step, in which the one-dimensional output is assigned into groups, representing the multidimensional components. This attitude is common practice in various applications. We prove that for piecewise stationary data, and when only second-order statistics are used, this form of separation is suboptimal. In particular, we obtain a closed-form expression for the MSE of this separation procedure, an expression which is based only on the model parameters. By comparing this expression with the MSE when the correct model is used, one can obtain the exact expected gain directly from the model parameters, without resorting to numerical simulations or Monte-Carlo trials.

Our analysis is verified through numerical experiments. In addition, we demonstrate the theoretical gain in the accuracy of component recovery in the presence of multidimensional components for several dependence scenarios.

Chair: prof. dr. ir.  S. Vanhuffel
Co- chair: Prof. Dr. Gunnar Naulaers

http://nieuws.kuleuven.be/node/11165

Early detection of epileptic seizures based on parameter identification of neural mass model

Speaker: Prof. Antoine Nonclercq, Ecole Polytechnique de Bruxelles, LIST Department, Brussels

Early seizure detection algorithms open new horizons to various applications. If reliable, such an algorithm could be implemented in an automatic warning system that alerts the patient of the occurrence of a seizure at an early stage. Besides, this system could also greatly benefit medical staff, allowing them to asses which specific functions may be impaired by a seizure and help them in localizing the source of the seizure activity. Another use for the algorithms is to trigger an imaging system (e.g., the injection of the radiotracer) for diagnostic purposes. Furthermore, electrical stimulation, when applied in an appropriate manner at seizure onset, has been reported to suppress spontaneous epileptiform activity. Using an early seizure detection algorithm, an ambitious system could therefore work in a closed loop to detect and terminate electrographic seizures.

Our group is developing a seizure detection method. It relies on parameter identification of a neurological model, which is attractive as its parameters can be explicitly related to neurological mechanisms. The occurrence of a seizure is evaluated by analyzing the shifts over time of key model parameters. The developed method also allows a better understanding of underlying electrical process during a seizure.

Programme
 

14h00-14h30:  Katrien Jansen

14h30-14h45: Carolina Varon

14h45-15h15: Alexander Caicedo

15h15-15h45: Hans De Clercq

15h45-15h55: Questions and Discussion

15h55-16h15: Break

16h15-16h30: Joseph Perumpillichira   ‘’temporal profiles of neonatal seizures”

16h30-17h00: Vladimir Matic

17h00-17h20: Ninah Koolen

17h20-17h30: Questions and Discussion

Programme
 

14h00-14h30:  Simon Tousseyn

14h30-14h40: Questions and Discussion

14h40-15h10: Borbola Hunyadi

15h10-15h20: Questions and Discussion

15h20-15h40: Coffee Break

15h40-16h10: Kris Cuppens

16h10-16h20: Questions and Discussion

16h20-16h50: Milica Milosevic

16h50-17h00: Questions and Discussion

• 14u00 - 14u30 :Vincent Verhaert
• 14u30 - 14u45 : Dorien Van Deun / Tim Willemen
•  14u45 - 14u50 : Q&A

• 14u50 - 15u05 : Devy Widjaja
• 15u05 - 15u35 : Carolina Varon Perez
• 15u35 - 15u40 : Q&A

• 16u00 - 16u20 : Dries Testelmans - expertise and needs within sleep lab
• 16u20 - 16u40 : Jan Van den Bulck - sleep related communication
• 16u40 - 17u30 : discussion and interaction

"Neurocognitive enhancement through exercise. Current approaches and applications"

Stefan Schneider (German Sport University Cologne).

The definition for health raised by the World Health Organization (WHO) includes physical and mental health. Today exercise science holds extensive knowledge about the adaptation of peripheral physiological systems to exercise (e.g. the hormonal, cardiovascular and musculoskeletal system).

Although the impact of exercise on mental fitness, cognitive performance and overall well-being has been extensively described in the recent decade, comparatively little is know about the underlying neurophysiological processes. This is mainly due to missing imaging possibilities as standardized imaging procedures, as positron emission tomography (PET) or functional magnetic resonance imaging (fMRI) are hardly applicable to health-orientated exercise settings. Nevertheless a deeper insight in the underlying neurophysiological parameters of exercise and their implications for neurocognition and emotional well-being are of utterly importance to a holistic understanding on how exercise promotes health.

Combined analysis of electrophysiological and hemodynamic signals

Katrien Vanderperren and Bogdan Mijović (K.U. Leuven, ESAT-SCD)

To study the cognitive processes going on in the human brain, both spatial and temporal information is needed. Most neuroimaging approaches, however, only possess high accuracy in one of these two domains and are thus not sufficient to fully assess the complexity of neural processes. For this reason, the integration of different modalities is becoming more and more popular.

One possible approach is the combination of electroencephalography (EEG) and functional magnetic resonance imaging (fMRI). This combination poses a series of challenges, ranging from recovering data quality to the fusion of two types of data of a completely different nature.

In this presentation several of these challenges will be addressed. We will show how the data are acquired and preprocessed and how the EEG and fMRI information can be fused.

For this fusion both two-dimensional methods like ICA as higher-dimensional decomposition methods can be used. Additionally, the potential benefit of exploiting sparsity and morphological diversity will be presented. The advantages and possibilities of these methods will be shown and discussed.

In medical practice, probabilistic prediction models are becoming increasingly important as supporting tools for diagnosis and decision making. Competing models for the same problem are often compared using the area under the ROC curve. However, this approach has serious limitations in a medical decision making context. For example, the difference in AUCs lacks clinical relevance and is relatively insensitive. Also, the AUC does not reflect performance of the model when it is used to classify patients to decide about treatment. Classification is achieved by using a sensible probability cut-off.

This presentation reviews alternative approaches that have been suggested in the literature, to finally arrive at the concepts of net benefit and test trade-off. Net benefit is an elegant measure to evaluate classification performance within a clinical utility framework. Models with higher net benefit are clinically more useful. The test trade-off combines the difference in net benefit between two models with the cost of using the models. The test trade-off represents the number of patients per extra true positive when using the model with highest net benefit instead of the other model. If this number is considered too high (i.e. fewer patients per extra true positive are desired), the additional benefit of the superior model does not compensate its additional cost such that it makes more sense to use the other model. These concepts allow for a more sensible comparison between competing models.

Weightlessness on Earth: from Newton and his apple to parabolic flights

André E Aubert

Labo experimentele Cardiologie

UZ Gasthuisberg, K.U.Leuven

According to legend, Isaac Newton was sitting under an apple tree, an apple fell on his head and he discovered the Universal Law of Gravitation. Gravity has shaped our world and governs life itself. Until 50 years ago it was not certain if humans could live without gravity. This possibility was proven by the historical orbital flight by Y Gagaryn on 12/4/1961 and since then more than 500 humans were in space under weightless conditions for periods up to 538 days.

Since flight costs for orbital flight are very elevated and it is restricted to a limited number of astronauts, many attempts have been made to create or simulate weightless conditions on Earth:

Simulations of microgravity:

1. Head out-of-water: subjects are immersed in water (at body temperature) for several hours.

2. Dry immersion: subjects spend time (a few days) on a kind of soft waterbed.

3. Head down bed-rest: subjects are in a bed under an angle of 6°, with head down for periods up to 180 days. They can only remain on their back or on their belly.

Real microgravity:

1. Drop tower: objects are in free fall in an evacuated tube for up to 4 s

2. Parabolic flights: a jet aircraft is free flying a parabolic trajectory (up to 9000 m), providing microgravity during 20-25 s for the aircraft and everybody and everything in it.

3. Sounding rockets: follow a parabolic free- flying trajectory (up to 100 km), providing 10-15 min microgravity.

Parabolic flight will be discussed in some depth and focus will be mainly on research on the effects of simulated microgravity on human physiology.

"Extraction features for myoclonic shock detection in video based on mean shift clustering for constructing motion tracks"

Kris Cuppens (K.U.Leuven, ESAT-SCD)

The golden standard in epileptic seizure detection makes use of video and EEG, where electrodes need to be attached to the scalp. This is not comfortable for the patient and medical staff is needed for the attachment of the electrodes. To make the monitoring of epileptic seizures in children feasible in a home situation, a detection system is needed that is easy to use and that is comfortable for the patient. We propose to detect nocturnal seizures with a motor component in patients by means of a single video camera. To this end we use a combination of optical flow and mean shift clustering to register moving body parts. These body parts are clustered in time into motion tracks. After extraction of seven features, related to amplitude, duration and direction of the motion, we carry out a first validation with a linear SVM.

Devy Widjaja (K.U. Leuven, ESAT-SCD)

Respiratory activity is usually measured with techniques like pneumography, spirometry and plethysmography. These methods are accurate, but have the disadvantage of interfering with natural breathing. It is however possible to overcome this problem by deriving the respiratory signal from other measurements that are influenced by respiratory activity. One of such measurements is the electrocardiogram (ECG), which is well-known to be affected by respiration in various ways.
On the one hand, respiration influences the heart rate by means of an acceleration during inspiration and a deceleration during expiration. This effect is known as respiratory sinus arrhythmia (RSA). Another effect comprises the chest movements and changes in the thorax impedance during the respiratory cycle due to the volume variations in the lungs, which influences beat morphology.
Based on these effects, several methods have been developed to extract a respiratory signal from the ECG.

Vladimir Matic (K.U. Leuven, ESAT-SCD)

An electroencephalogram (EEG) is present in research and in clinical practice for the last 70 years. However, the state of the art analysis and interpretation of the EEG is still obtained by visual examination of an experienced neurophysiologist. This is mainly due to the complex morphology of the EEG signal. In order to develop and improve automated algorithms for the analysis of the EEG signal an efficient parameterization is necessary. Methods should also provide high time­­-frequency resolution and sparse approximation. For that purpose, application of wavelets and local cosine basis has been investigated. Furthermore, in order to achieve higher sparsity, decomposition on Gabor atoms have been applied with basis pursuit algorithms. Following this approach we get better insight into the morphology of the EEG signals that can help us in classification and analysis tasks.

10:00-11:00 "The use of wavelet transform in Cerebral Autoregulation assessment"

Alexander Caicedo (K.U. Leuven, ESAT-SCD)

Cerebral autoregulation is defined as the capacity of the brain to keep a constant cerebral blood flow (CBF) despite the changes in the cerebral perfusion pressure. The most important forms of brain injury in premature infants are presumably partly caused by disturbances in this mechanism. As changes in cerebral intravascular oxygenation (HbD), regional cerebral oxygen saturation (rSO2), and cerebral tissue oxygenation (TOI), all recorded with Near-Infrared Spectroscopy (NIRS), reflect changes in cerebral blood flow (CBF), impaired autoregulation can be measured by studying the concordance between HbD, rSO2 or TOI and the mean arterial blood pressure (MABP), assuming no changes in oxygen consumption, arterial oxygen saturation (SaO2), and in blood volume. Up to now, there is no strong evidence that links impaitred cerebral autoregulation (assessed by classical methods as correlation, coherence and transfer function analysis) and clinical outcomes in neonates. This result can be partly attribuited to the highly time dependant behaviour of cerebral autoregulation, as classical methods take into account the overall signal dynamics possibly hidden important localized phenomena. In order to have a better understanding of the behaviour of these mechanism in time wavelet transform is used.

11:00-12:00 "Automatic Seizure Detection incorporating structural information"

Bori Hunyadi (K.U. Leuven, ESAT-SCD)

Traditional seizure detection algorithms act on single channels ignoring the synchronously recorded, inherently interdependent multichannel nature of EEG. However, the spatial distribution and evolution of the ictal pattern is a crucial characteristic of the seizure. Two different approaches aiming at including such structural information into the data representation will be presented. Their performance is compared to the traditional approach both in a simulation study and a real-life example, showing that spatial and structural information facilitates precise classification.

"Simultaneous facial EMG & fMRI recordings"

Milica Milosevic (K.U. Leuven, ESAT-SCD)

Jan Luts (K.U.Leuven, ESAT-SCD)

This presentation introduces a technique which extends the least squares support vector machine (LS-SVM) classifier to a mixed effects LS-SVM classifier which can handle longitudinal data. The mixed effects LS-SVM model contains a random intercept and allows to classify highly unbalanced data, in the sense that there is an unequal number of observations for each case at non-fixed time points. This semi-parametric approach consists of a regression modeling and a classification step based on the obtained regression estimates. Regression and classification of new cases are performed in a straightforward way by solving a linear system. It is demonstrated that the methodology can be generalized to deal with multi-class problems and can be extended to incorporate multiple random effects. The technique is illustrated on simulated data sets and real-life problems concerning human growth.

Everybody is confronted with the diagnosis of cancer during their life. If not personally, it might be a family member or a friend. The diagnosis of cancer creates a feeling of fear. Fear for the unknown, fear for the therapy, fear to die. Thanks to the knowledge of clinical doctors the patients can be informed on their disease and can be treated adequately. In order to decide on treatment options, estimation of risk of survival is crucial. Although chemotherapy might reduce the risk on metastatic tumors, the therapy in itself might be harmful. Deciding which patients might benefit from a specific type of therapy in a way that the benefit is higher than the negative effects of the therapy, is of major importance in clinical practice. These questions are answered by means of survival analysis. By studying patient variables and their relation with survival time (time to relapse, time to death, time to healing) an estimate of the patient specific risk can be calculated. In addition, this type of models allows to estimate the effect of different types of treatment. The information which is provided to clinicians in this way can help them making a more adequate treatment choice.

Survival data are most often analyzed by parametric and semi-parametric models. Although these models perform well in general, some drawbacks remain. The most popular model for survival analysis is the proportional hazards model (ph). The name of this model reveals that hazards for observations with different variable values are assumed to be proportional. However, this assumption is not always realistic. Secondly, the model assumes a linear parametric form of the variables.

To overcome these drawbacks, this research presents a new mathematical model for the analysis of survival data. The largest threshold in developing survival models is the occurrence of censoring in the data. Censoring indicates that not all survival times are observed exactly. Some patients might not experience the event under study during the study time, others might be lost to follow-up. To deal with this issue, the survival problem is reformulated as a ranking problem, where only pairs of observations for which the event order is known, are taken into account. This principle is combined with the methodology of support vector machines. However, the maximal margin principle is replaced by minimizing the Lipschitz constant.  This model is further explored by considering different adaptations, such as L2 instead of L1 norms and regression constraints instead of ranking constraints. This research revealed that the use of equality constraints is less appropriate in analyzing survival data since the information provided by censored data can not completely be included. In addition, it is shown that the inclusion of regression constraints improves the performance significantly.

A second achievement of this research is that the proposed method appears to be very promising when handling high dimensional data. The last decade, more and more research was done with regard to genomics, proteomics and other “omics”, resulting in high dimensional data. The problem with conventional techniques is that the estimates of the effects of each variable becomes less reliable when more and more variables are included. The presented method overcomes this problem by solving the optimization problem in the dual space.

Although our experiments illustrate that the proposed model results in well performing models, the chance that this model will be used in clinical practice is very small. The problem with the clinical use of non-parametric models is that they are black-box models. As a result, clinicians can not be provided with information on how the model estimates a certain risk, making it hard for clinicians to trust the model. In addition, this type of models can not be used to search for new predictive and prognostic markers since one does not know which variables contribute to the result. These major drawbacks are solved in the interval coded scoring index (ics). Although this new model starts from the model explained above, the results are highly interpretable and easily applicable in clinical practice.  The basic idea of the ics is that variables can be divided into a relatively small number of intervals, in which the effect on the risk of the event remains the same. In addition, it is assumed that the effects of all variables can be summed up to become the final risk. Classification and prognostic models developed by means of this new methodology can be presented as a questionnaire or by means of color bars, which makes application possible within software applications as well as on paper.

The MALDI-TOF mass spectrometry is a useful technique for the analysis of proteomics. To compare samples from different biological conditions, they are often processed in the same spectrum in order to avoid between-spectra variability. The enzymatic 18O-labeling is an often used labeling technique to separate samples of different conditions, appearing in the same spectrum. A limitation of the technique is the possibility of an incomplete labeling, which may result in biased estimates of the relative peptide abundance. To address this issue, we propose a set of Markov-chain-based regression models. The model was initially implemented based on the summary statistics of the original settings of the data, termed as the stick representation. It is then extended by incorporating heteroscedasticity (mean-dependent variance function), random effects (for the assessment of between-spectra technical or biological variability). The models were formulated in both frequentist and Bayesian frameworks. Finally, the model was extended for the application of the original settings of the data set, by using an appropriate shape function. To assess the application values of the implemented models, they were applied to a real-life case study. In the mean time, several simulation studies were also conducted to investigate the statistical performances.

Symposium ‘Cardiovascular monitoring’

13.30-14.30 Prof. Dr. Ir. Sergio Cerutti

Cardiovascular Variability Signals: Towards a Quantitative Assessment of the Complexity of Autonomic Controlling Systems

14.30-15.00 Prof. Dr. Ir. Bob Puers

Wearable ECG monitoring systems

15.00-15.30 Prof. Dr. Andre Aubert

Control of heart rate and physiological link to heart rate variability. Application to human space flight.

15.30-16.00 Prof. Dr. Gunnar Naulaers

Use of Physiological Monitoring in Neonatal Intensive Care

16.00-16.30

Coffee break

PhD defense

16.30-18.30

PhD defense Steven Vandeput

18.30-20.00

Reception

Location: Arenberg Castle, Kasteelpark Arenberg 1, 3001 Heverlee (Leuven)

Auditorium 01.07

Free participation, but registration obligatory

Email: steven.vandeput@esat.kuleuven.be

This interdisciplinary course is part of the thematic training of the Leuven Arenberg Doctoral School Training Programme and supported by IAP BCRYPT and LICT. The course is mainly aimed at Ph.D. students from all disciplines (either from the K.U.Leuven or from other universities,) but also open to undergraduate students, post-docs, people working in industry, or anyone else interested on the topic.

The course will provide an overview of various aspects of privacy from the technical, legal, economics, and social science perspectives. While the broad focus of the course is on privacy in electronic services, this year’s edition of the course will have a special focus on social networks.

When

·            Wednesday, June 23, from 09:30 to 17:30

·            Thursday, June 24, from 09:00 to 18:00

Where

Computer room at the Mediacentre (Faculty of Social Sciences)

Speakers

The course will last two days and consist of eight lectures. The lecturers include five speakers from different departments and faculties in K.U.Leuven and an invited speaker:

·            Prof. Alessandro Acquisti, (Carnegie Mellon University, USA)

·            Prof. Bettina Berendt, Computer Science (K.U.Leuven)

·            Dr. Claudia Diaz, Electrical Engineering (K.U.Leuven)

·            Dr. David Geerts, Faculty of Social Sciences (K.U.Leuven)

·            Seda Gürses, Electrical Engineering / Computer Science (K.U.Leuven)

·            Eleni Kosta, Faculty of Law (K.U.Leuven)

Registration

·            The course is free of charge, but attendees are required to register by sending an email to claudia.diaz@esat.kuleuven.be

·            The registration deadline is: Tuesday, June 15

If you have any questions or would like to know more information please send an email to claudia.diaz@esat.kuleuven.be.

Programme

Wednesday, June 23

09:30 Introduction (Claudia Diaz)

10:30 Coffee break

11:00 Overview of Privacy Enhancing Technologies (PETs)  (Claudia Diaz)

12:30 Lunch break

14:00 Exploring European data protection: From social networks to cookies (Eleni Kosta)

15:30 Coffee break

16:00 Privacy and Web mining (Bettina Berendt)

17:30 End

Thursday, June 24 

09:00 To share or not to share - a user's perspective on privacy in social networks (David Geerts)

10:30 Coffee break

11:00 Privacy Concerns and Information Disclosure: An Illusion of Control Hypothesis (Alessandro Acquisti)

12:30 Lunch break

14:00 Privacy, Requirements Engineering and Online Social Network Services (Seda Gürses)

15:30 Coffee break

16:00 Predicting Social Security Numbers From Public Data (Alessandro Acquisti)

17:30 Discussion speakers and participants

18:00 end

"Approaches to solving High Density surface EMG decomposition problem"

Ivan Gligorijevic (K.U. Leuven, ESAT-SCD)

Neuromuscular disorders can be of very diverse nature and severity, but they have in common that they affect motor units. Motor units are the functional units of the peripheral motor system. They consist of a single motor neuron together with the (20 to 2000) muscle fibres that this neuron innervates. In voluntary contractions, each motor unit has its own pattern of activation, with firing moments that are only weakly related to those of other motor units. Recordings of single motor unit potentials in patients reflect the underlying pathology and can hence be employed for diagnostic and monitoring purposes. Over the past decade, a noninvasive alternative to intramuscular “needle-electrode” measurements has been developed. This new approach uses an array of up to 128 surface electrodes, densely spaced over the skin above the muscle of interest (high-density surface EMG or HDsEMG). Whereas single-channel surface EMG recordings lack the specificity that is required to identify single motor unit potentials, the additional, mostly spatial information that can be derived with HDsEMG does appear to make this feasible. The problem is how to positively extract individual muscle units and their firing “signatures” from intense, often overlapping observations. Ways to approach this problem differ: one is to apply clustering techniques and discard the overlaps, others include solving optimization problem on the known “vocabulary” of individual muscle units (deals with overlaps as well) and there is also approach connected to solving convolutive-ICA (independent component analysis) problem.

"Development of an epileptic seizure detection system"

Borbala Hunyadi (K.U. Leuven, ESAT-SCD)

Epileptic seizure detection and development of a seizure onset alarm system has been of interest since already a few decades. In spite of the numerous attempts, a perfect solution has not been found yet. This is due to mainly two factors: the inter-patient variability of seizure characteristics provoke a trade-off between sensitivity and specificity; besides, artefacts contaminating the measurement data cause false detections or cover ictal patterns causing missed seizures.

During the development of the seizure detection algorithm I concentrate on the visually appearing characteristics, on which neurologists also rely when reading the EEG data. I will present the main step of the detection procedure, and show some considerations on how to overcome the difficulties posed by the various artefacts present on the data.

"Application of data-driven techniques in integrated EEG-fMRI analyses"

Katrien Vanderperren and Bogdan Mijovic (K.U. Leuven, ESAT-SCD)

Multimodal approaches are of growing interest in the study of neural processes. To this end much attention has been paid to the integration of electroencephalographic (EEG) and functional magnetic resonance imaging (fMRI) data. EEG and fMRI are not only among the most widely employed techniques in neuroscientific research but also offer a perfect complementarity when it comes to their temporal and spatial resolution.
The integrated analysis of EEG and fMRI is not straightforward. First, the simultaneous acquisition of EEG and fMRI causes severe artifacts in the data. Second, a serious challenge is posed by the different nature of both types of data.
In this presentation we will therefore show the advantages of data-driven techniques at different stages of this EEG-fMRI processing. We will explain the entire analysis process starting from the simultaneous acquisition of EEG and fMRI until the spatiotemporal information extracted from them.

ANNOUNCEMENT of  a K.U.Leuven Seminar, Campus ARENBERG,

Prof.  Ingrid DAUBECHIES
(Princeton university, USA)

“Surfing with Wavelets”

Tuesday May 11, 2010 from 11h00 to 12h00

Thermotechnisch Instituut, Kasteelpark Arenberg 41, 3001 HEVERLEE
Aula Van de Tweede Hoofdwet (TI 01.0002)

ABSTRACT:

Wavelets are used in the analysis of sounds and images, as well as in many other applications. The wavelet transform provides a mathematical analog to a music score: just as the score tells a musician which notes to play when, the wavelet analysis of a sound takes things apart into elementary units with a well defined frequency (which note?)  and at a well defined time (when?). For images wavelets allow you to first describe the coarse features with a broad brush, and then later to fill in details. This is similar to zooming in with a camera: first you can see that the scene is one of shrubs in a garden, then you concentrate on one shrub and see that it bears berries, then, by zooming in on one branch, you find that this is a raspberry bush. Because wavelets allow you to do a similar thing in more mathematical terms, the wavelet transform is sometimes called a "mathematical microscope".

Wavelets are used by many scientists for many different applications.  Outside science as well, wavelets are finding their uses: wavelet transforms are an integral part of the image compression standard JPEG2000.

The talk will start by explaining the basic principles of wavelets,  which are very simple. Then they will be illustrated with some examples, including an explanation of image compression.

BRIEF BIOGRAPHY:

Ingrid Daubechies received both her bachelor's and Ph.D. degrees (in 1975 and 1980) from the Free University in Brussels, Belgium.  She held a research position at the Free University until 1987. From 1987 to 1994 she was a member of the technical staff at AT&T Bell Laboratories, during which time she took leaves to spend six months (in 1990) at the University of Michigan, and two years (1991-93) at Rutgers University. She is now at the mathematics department and the Program in Applied and Computational Mathematics at Princeton University.  She was awarded a Leroy P. Steele prize for exposition in 1994 for her book Ten Lectures on Wavelets. From 1992 to 1997 she was a fellow of the John D. and Catherine T. MacArthur Foundation. She is a member of  the American Academy of Arts and Sciences, the American Mathematical Society, the Mathematical Association of America, the Society for Industrial and Applied Mathematics, and the Institute of Electrical and Electronical Engineers. She is married and has two children.

ORGANIZERS:

The seminar is organized by the Departments of Electrical Engineering, Mathematics and Computer Science on the occasion of the appointment of Prof. Daubechies as International Francqui Professor at the Free University in Brussels (January-June 2010).

"The influence of line-shape and baseline on quantification of/in vivo/magnetic resonance spectroscopy (MRS) signals"

 Maria Isabel Osorio (K.U. Leuven, ESAT-SCD)

To study the influence of baseline and lineshape distortions on /in vivo/ Magnetic Resonance Spectroscopy (MRS) signals, we analyzed spectra obtained from epileptic rats. MRS signals are commonly affected by magnetic field inhomogeneities and tissue heterogeneity, distorting the ideal Lorentzian lineshape; additionally, short-T2 components (macromolecules) also influence the normal baseline of spectra. Therefore, to carefully quantify these signals, we use the quantification method AQSES which accounts for the baseline using splines and we included an algorithm to correct for lineshape distortions using the self-deconvolution method that calculates a common lineshape in the spectrum under analysis.

"On the use of a clinical kernel in survival analysis"

Vanya Van Belle (K.U. Leuven, ESAT-SCD)

Clinical datasets typically contain continuous, ordinal, categorical and binary variables. To model this type of datasets, kernel based methods are generally used in combination with a linear kernel. However, this kernel has some disadvantages, which were tackled by the introduction of a clinical kernel [Daemen et al., 2009].  This seminar shows the results of experiments, comparing the test performance of kernel based survival models using linear and clinical kernels. In addition, some important properties of the clinical kernel are investigated.  In first instance, the question is raised whether the clinical kernel satisfies Mercer's condition, a necessary condition for kernels used in kernel based models.  Secondly, we wonder whether the clinical kernel is still a linear one.  If not, how fair is the comparison with the linear kernel? 

"EEG and ERPs in the functional investigation of cognitive processes"

Anna M. Bianchi

Dep of Biomedical Engineering, Politecnico di Milano, Italy

Abstract

It is now 80 years that Hans Berger published his first paper on EEG. In the following decades EEG became an irreplaceable tool in clinics and for the functional investigation of the brain. The EEG recorded on the scalp is generated by the synchronized activity of millions of cortical neurons, thus processing methodologies able to quantify different kinds of synchronizations are employed for the description of the cortical functions. Different EEG correlates have been associated to cognitive functions and different processing techniques have been proposed in order to highlight the related information. Filtering and time frequency analysis are used to describe the modulation in time of the different EEG rhythms during attention tasks; measures of synchronization among different cortical areas are achieved through linear and non-linear methodologies.

Synchronization of neural populations following sensorial or cognitive stimulations is measured through evoked potentials and event related potentials. Advanced processing techniques are employed for single sweep analysis and for the calculation of templates.

In the more recent years the information coming from EEG is also used for a more focused and detailed analysis of fMRI or NIRS data, and the integration of the different functional techniques allow to combine high time resolution (EEG), with high spatial resolution (fMRI, NIRS).

"Spike detection and analysis in deep brain recordings"

Ivan Gligorijevic (K.U. Leuven, ESAT-SCD)

Deep brain stimulation has been used for years to treat some medical disorders like Parkinson, OCD, etc… But it has been done mostly by trial and error. In order to understand effects and needs for stimulation better a closed loop system is needed between stimulation and recording from roughly the same sight in brain. First step towards this is being able to process and analyze signals you record. These signals are actually so called spike-trains, which are extracellular observation of neural activity. Being able to distinguish between different neurons and analyze their spike activity during time is a crucial and very challenging task about which many papers have been published but without definite conclusion about best or even optimal solution. New signal processing techniques are finding way into this area daily.

"Differentiation between brain metastases and glioblastoma multiforme based on MRI, MRS and MRSI"

Jan Luts (K.U. Leuven, ESAT-SCD)

Brain metastases and glioblastoma multiforme are the most aggressive and common brain tumours in adults, but they require a different clinical management. Anatomical magnetic resonance imaging (MRI) or clinical history, cannot always clearly distinguish between them. This study describes and verifies the use of magnetic resonance spectroscopy (MRS) and magnetic resonance spectroscopic imaging (MRSI) in combination with MRI for differential diagnosis of glioblastomas and metastases. Feature selection methods are applied to the magnetic resonance (MR) spectra of 121 patients and relevant features are detected. Different classification methods are used to distinguish glioblastoma multiforme and metastasis based on the single-voxel MR spectra, but no reliable differentiation is obtained: the accuracy varies from 50 to 78%. Next, MRSI and MRI data from 10 patients (5 glioblastomas, 5 solitary metastases) are used for differentiation purposes. The combination of multivoxel MR data and MRI data suggests a more clear differentiation between glioblastoma multiforme
and brain metastasis. The results are visualized based on nosologic images, which are generated by including spectroscopic information in the segmented MR image. The methodology offers a new way that may support clinicians in decision making.

"Advances in simultaneous EEG-fMRI acquisition and analysis"

 Katrien Vanderperren (K.U. Leuven, ESAT-SCD)

Several techniques are available to study cognitive processes in the human brain but none of these is able to offer both a high temporal and spatial accuracy. Therefore the combination of electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) has been suggested as a tool to overcome this problem. The two proposed techniques are complementary in the sense that EEG has a high temporal resolution while fMRI has a very good spatial resolution. For the integration of EEG and fMRI, their simultaneous acquisition is essential but it causes severe artifacts on the EEG. In this presentation, these quality aspects are being discussed and the techniques used to remove the artifacts are shown. Moreover, several possible approaches to integrate the information coming from both modalities, are being presented.

"Neonatal seizure localization using the CP-decomposition"

Wouter Deburchgraeve (K.U. Leuven, ESAT-SCD)

Neonatal seizures are an important sign of brain damage in the neonate. Their localization in the brain is important for diagnostic and prognostic implications. Multichannel EEG recordings allow topographic localization of seizure foci. In this presentation, two EEG-based algorithms are introduced for an automatic and objective determination of the seizure location in the neonatal brain as it is reflected on the scalp. Each algorithm extracts the electrical potential distribution of the seizure over the scalp using the higher-order canonical decomposition, also referred to as the CP model. The resulting localization is compared with that of visual analysis of the EEG by an experienced clinical neurophysiologist, and the similarity of both analysis is shown.

"MRS lineshape determination applying AQSES"

Maria Isabel Osorio Garcia (K.U. Leuven, ESAT-SCD)

In vivo Magnetic Resonance Spectroscopy (MRS) is a noninvasive technique used for monitoring biochemical information of metabolites in tissue. This allows comparison and quantification of normal and abnormal tissue. Quantification methods like AQSES, QUEST and LCModel allow the determination of metabolites, however,  it is necessary to take into account different distortions that commonly affect the natural damping of MRS signals and thus the accurate quantification results. The lineshape of MRS signals can be influenced by inhomogeneities in the magnetic field and heterogeneities within the tissue, for instance at the interface between brain and bone tissue. Therefore, we use the quantification method AQSES by applying a lineshape estimation method which require setting certain parameters/hyper-parameters necessary for accuracy.

"Quantification and classification of in vivo MRSI data using multimodal information"

Anca Croitor (K.U. Leuven, ESAT-SCD)

In this presentation we show how combining multimodal information coming from magnetic resonance imaging (MRI), magnetic resonance spectroscopic imaging (MRSI) and/or high resolution magic angle spinning (HR-MAS) can improve the performance of quantification and/or classification of MRSI data.  During an MRSI acquisition, Magnetic Resonance spectra are measured in a grid of voxels and we propose a quantification method  for MRSI data that exploits spatial prior knowledge. Different ways of exploiting spatial information are presented, discussed and compared. Further we analyze whether, by using all the information available for a tumor type, even when this information is acquired in different centers or is obtained by different NMR techniques (MRI, MRSI and HR-MAS), it is possible to improve the performance of the classifier in terms of accuracy and reliability. This approach can be extended to any pattern recognition method/system, which makes use of a learning procedure.     

"The power of computer-aided biosignal processing in neonatal brain diagnostics"

Sabine Van Huffel (K.U. Leuven, ESAT-SCD)

This lecture give an overview of classical as well as advanced biosignal processing methods for monitoring neonatal brain and detecting risk situations for brain damage. Our study is focused on /brain signals (EEG, NIRS) of (preterm) newborns at risk for brain //damage due to shortage in the oxygen supply (= asphyxia). First of all, we compare 3 different methods (correlation, coherence, partial coherence) for detection of impaired cerebral autoregulation. / Impaired cerebral autoregulation may lead to oxygen insufficiency and hence brain damage.
Brain damage leads to seizures which need to be treated to minimize neurodevelopmental disorders. We show how to  automate neonatal seizure monitoring  including seizure detection, ECG artefact removal and seizure localization.

"Empirical Mode Decomposition and other Techniques for Decomposing Single Channel Signals"

Bogdan Mijovic (K.U. Leuven, ESAT-SCD)

     The Empirical Mode Decomposition (EMD) is a novel signal analysis tool which is able to decompose any complicated time series into a set of spectrally independent oscillatory modes called Intrinsic Mode Functions (IMFs). The advantage of EMD compared to Wavelet analysis is that EMD is able to deal with non-stationary and non-linear data. While Wavelets and other signal decomposition techniques tend to map the signal space onto a space spanned by a predefined basis, EMD is data driven algorithm which means that it decomposes the signal in a natural way where no a priori knowledge about the data embedded in the signal of interest is needed.

            On the other hand, Independent Component Analysis (ICA) is well known and well established blind source separation technique which can separate a set of signals into statistically independent sources (components). However, this is possible only when the number of electrodes (channels) is larger then or equal to the number of sources. Unfortunately with the single channel data this is possible only in the trivial case, where number of sources is one, i.e. the signal is coming from only one source.

            The joining of the two methods in order to achieve single channel signal decomposition into its independent components will be presented. Results of different simulations and comparison with other methods will be shown.

"Advances in Total Least Squares with multiple right-hand sides"

Diana Sima (K.U. Leuven, ESAT-SCD)

We consider linear approximation system AX=B and focus on the Total Least Squares problem, which assumes that both given matrices A and B are noisy. The Total Least Squares solution X should exactly solve a nearby compatible system (A+E)X=B+G and the corrections E and G should be as small as possible. We investigate necessary and sufficient conditions for existence of the Total Least Squares solution in the multiple right-hand sides case. Our complete classification reveals that the TLS solution can in some situations be different from the output returned by the classical TLS algorithm. We also discuss the concept of core problem in a linear algebraic system, which is a minimally dimensioned problem that contains all necessary and sufficient information for solving the original system, and focus on an extension of the core problem to systems with multiple right-hand sides.

"Overview of numerical methods for best multilinear rank approximation of tensors"

Mariya Ishteva (K.U. Leuven, ESAT-SCD)

Multi-way arrays are used in many application fields, such as statistics, signal processing and scientific computing. Efficient and reliable algorithms for manipulating higher-order tensors are thus required. In this talk, we focus on the best rank-(R1,R2,R3) approximation, which can be used for dimensionality reduction. This problem is more complicated than the best rank-R approximation of a matrix. Truncation of the higher-order singular value decomposition only yields a good starting point for iterative algorithms.

In this talk, we present three new algorithms. We first formulate the solution of the optimization problem as a zero of a well-chosen function F. Newton’s method cannot be applied directly on F because of a symmetry property by the action of the orthogonal group. We propose a geometric Newton method that removes the symmetry by working on a quotient manifold. We illustrate the algorithm’s fast convergence. For the other two algorithms, we express the tensor approximation problem as the minimization of a cost function on a product of three Grassmann manifolds. We develop both a Riemannian trust-region and a conjugate gradient methods. We show some applications.

"Detection of a mental load in physiological signals"

 Joachim Taelman (K.U. Leuven, ESAT-SCD)

Stress is a huge problem in today’s society. It leads to muscle overload, depression,… Besides the psychological origin of stress, the phenomenon affects several physiological systems in the body: Neurological, muscular, hormonal, respiratory,… In this study, we mainly focus on the changes in the muscular and the cardiac system via the electrophysiological signals of the Trapezius muscle and the heart. The results of the influence of a mental task on the variability of the heart and the respiration are presented in this seminar. 

"Linear and nonlinear Heart Rate Variability analysis and its applications"

Steven Vandeput (K.U. Leuven, ESAT-SCD)

"Additive survival least squares support vector machines"

Vanya Van Belle (K.U. Leuven, ESAT-SCD)

This talk discusses a new survival modelling technique based on
Least Squares Support Vector Machines.  We
propose the use of a least squares support vector machine
combining ranking and regression.  The advantage of this
kernel-based model is threefold: (i) the problem formulation is
convex and can be solved conveniently by a linear system, (ii)
non-linearity is introduced by using kernels, componentwise
kernels in particular are useful to obtain interpretable results,
(iii) introduction of ranking constraints makes is possible to
handle censored data.    In an experimental set-up, the model is
used as a preprocessing step for the standard Cox proportional
hazard regression by estimating the functional forms of the
covariates, or for any linear survival model. The proposed model
is compared with different survival models from the literature on
the German Breast Cancer Study Group data. 

"Cardio-respiratory system during sleep: application for decision support systems"

Speaker: Martin Mendez,

Dept. of Biomedical Engineering,

Politecnico di Milano, Pzza. Leonardo da Vinci 32, Italy

In the past decades, the progress in mathematical modeling and the availability of computational resources paved the way to the resolution of relevant and time consuming problems. These advances allow the analysis of large amounts of data and the development of decision support systems (DSS) in order to reduce repetitive and tiring works to human experts. In many different clinical disciplines, DSS may aid to establish feasibility of measurements and develop automatic screening based on more data in a shorter time. In particular in sleep medicine, it is usual practice to perform the sleep staging procedure through the visual scoring of polygraph traces, that include many signals (electroencephalography EEG, electromyography EMG, electrooculogram EOG, respiration, and others) recorded along the whole night. Such a polysomnography (PSG) evaluation is the gold standard procedure for sleep staging and in general for the diagnosis of the main classical sleep disorders. However, PSG presents some inconvenient such as the need for some dedicated equipment, dedicated sleep centers, specialized and expert technical personnel. Due to the reduced number of sleep centers, sleep diagnosis is a very expensive procedure and sometimes the waiting list for the patients is very long. As a consequence, many relevant sleep disorders remain underestimated and untreated with many drawbacks for the health. All these considerations lead to develop automatic procedures for the analysis of the sleep.

It has been observed that some peripheral measures, such as Heart Rate Variability (HRV), blood pressure and respiratory activity, present specific oscillatory patterns that are modulated in relation to the different sleep stages. For instance, during NREM sleep (non rapid eye movement), HRV presents spectral components well concentrated around the respiratory frequency. While during REM (rapid eye movement) sleep, HRV spectrum presents low frequency components, which seem to be linked with the instability in the EEG cerebral waves. Specific mathematical models could capture these typical variations and feed into a classifier in order to define a DSS to obtain a fast sleep screening in any site such as it could be home.

HRV offers advantages such as high signal-to-noise ratio and it is of very simple acquisition. These characteristics make HRV a very interesting signal for the development of DSS. In addition, the spectral components of the HRV are strongly related to the Autonomic Nervous System (ANS), thereby broadening its use. The behavior of the HRV has been largely explored during sleep and many studies have been focused on its use for fast screening of sleep apnea. The results obtained by those studies present high classification performance between normal and pathologic sleep.

During the lecture some examples of signal processing and pattern recognition techniques together with the cardio-respiratory system will be presented in order to show the potentiality for developing clinical DSS.

"Empirical Mode Decomposition – The New Way for Analyzing non-Linear and non-Stationary Signals"

Empirical Mode Decomposition (EMD) is new technique for analyzing nonlinear and non-stationary data. It decomposes any complicated data series into a finite and often small number of mono-component functions called `Intrinsic Mode Functions' (IMF’s). These functions are based on local properties of the signal, and therefore they are able to provide us with meaningful instantaneous amplitudes and instantaneous frequencies as functions of time, allowing us to identify imbedded structures in the signal. This decomposition method is adaptive, and, therefore, highly efficient. Since the decomposition is based on the local characteristic time scale of the data, it is applicable to nonlinear and non-stationary processes. The method itself will be presented together with the advanced, noise assisted method for extracting the IMF’s – Ensemble Empirical Mode Decomposition (EEMD). A number of advanced algorithmic variations for computing instantaneous amplitude and instantaneous frequency will be discussed. Finally, application of the method on Heart Rate (HR) signal for detecting sleep apnea is shown and briefly discussed.

"Automated analysis of neonatal EEG"

Wouter Deburchgraeve (K.U. Leuven, ESAT-SCD)

"Removal of artifacts in simultaneous EEG and fMRI experiments"

Katrien Vanderperren (K.U. Leuven, ESAT-SCD)

"Physiological stress detection in EMG signals"

Joachim Taelman (K.U. Leuven, ESAT-SCD)

"Imposing independence constraints to the CP model"

Maarten De Vos (K.U. Leuven, ESAT-SCD)

"Best rank-(R1,R2,R3) tensor approximation algorithms"

Mariya Ishteva (K.U. Leuven, ESAT-SCD)

Abstract:
Higher-order tensors are generalizations of vectors (order 1) and matrices (order 2) to order 3 or higher. They have various application areas, such as biomedical engineering, image processing, and signal processing.
In this talk, we first consider some basic properties of a higher-order tensor. Contrary to the matrix case, the tensor best rank-(R1,R2,…,RN) approximation cannot be computed in a straightforward way. In the second part of the talk, we present the higher-order orthogonal iterations algorithm and derive two new algorithms for computing the best rank-(R1,R2,…,RN) approximation of a tensor, based on the trust-region and conjugate gradient methods on manifolds. We touch on some of the applications.

Support Vector Machines for Survival Analysis

Vanya Van Belle (K.U. Leuven, ESAT-SCD)

Traditional statistical techniques as based on Cox' proportional hazard model or the accelerated failure time model focus on explicitly modeling the underlying probabilistic mechanism of the event under study. Machine learning techniques as based on Support Vector Machines (SVM) on the contrary take a rather different perspective.  Their main focus is to learn a predictive rule which will generalize well to unseen data. SVMs have been used for prognostic reasons by reformulating the survival problem as a classification problem, dividing the time axis in a number of predefined intervals or classes.  We propose a different approach by maximizing the concordance index between observed event times and estimated ranks of event occurrence.  Additive models are used to improve interpretability of the models.

Combined MRI and MRSI segmentation and classification

 Jan Luts (K.U. Leuven, ESAT-SCD)

A nosologic image summarizes the existence of different tissues in a single image which makes it easy to interpret for clinicians. Each voxel or pixel of the image is coloured according to the histopathological class it belongs to. In [1] and [2] MRI was combined with MRSI information to create nosological images. However, each voxel or pixel was treated independently, not including any spatial information. To exploit neighbourhood information in MRSI, canonical correlation analysis was proposed in [3]. In the present study we apply more advanced methods from image processing and pattern recognition to segment and classify brain tumours, thereby including spatial information. MRSI and MRI data are combined to produce higher resolution nosologic images. Furthermore, class probabilities are calculated for the segmented tumour region.

[1] F.S. De Edelenyi et al., Nat Med 6 (2000), pp. 1287-1289
[2] A.W. Simonetti et al., Anal Chem 75 (2003), pp. 5352-5361
[3] T. Laudadio et al., Magn Reson Med 54 (2005), pp. 1519-1529

Numerical Noise Titration Technique and its Applications in HRV

 Steven Vandeput (K.U. Leuven, ESAT-SCD)

Abstract:
Heart Rate Variability (HRV) measurements are used as markers of the autonomic modulation of the heart rate. Standard (linear) time and frequency domain methods of HRV are well described by the Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology, but in the last decades, new dynamic methods of HRV quantification have been used to uncover apparent nonlinear fluctuations in heart rate. These nonlinear variations would enable the cardiovascular system to respond more quickly to changing conditions. In this talk, an overview of the existing nonlinear HRV measures will be given. In particular, the recently developed numerical noise titration technique will be explained, which provides a highly sensitive test for deterministic chaos and a relative measure for tracking chaos of a noise-contaminated signal in short data segments. That technique was applied on different datatypes of which the results will be discussed in the second part of the presentation.

Truncated Total Least Squares

Diana Sima (K.U. Leuven, ESAT-SCD)

Abstract:
The method of truncated total least squares is an alternative to the classical truncated singular value decomposition used for the regularization of ill-conditioned linear systems. Truncation methods aim at limiting the contribution of noise or rounding errors by cutting off a certain number of terms in an expansion such as the singular value decomposition. To this end a truncation level k must be carefully chosen. The truncated total least squares solution becomes more significantly dominated by noise or errors when the truncation level k is overestimated than the truncated singular value decomposition solution does. We propose a modified generalized cross validation method, which combined with the truncated total least squares method performs better than the classical generalized cross validation combined with the truncated singular value decomposition.

Near infrared imaging brings new light in brain research

PD Dr. Martin Wolf, Ph. D. Engineering and lecturer

Clinic of Neonatology, University Hospital Zurich, Frauenklinikstr. 10, 8091 Zurich, Switzerland

Near-infrared imaging (NIRI) is a quickly growing method to non-invasively study human brain tissue using near infrared light, which penetrates tissue several cm deep. The method is appreciated by patients and researchers, because it is quantitative, measures continuously, is painless, can be used at the bedside, is relatively inexpensive and can easily be combined with other modalities such as e.g. fMRI and EEG. Based on established physical models light absorption and scattering of tissue can be measured, which in turn yields important physiological parameters (e.g. oxygenation) and allows to monitor the function of biological tissue. Using multiple wavelengths the concentration of constituents of tissue such as oxyhemoglobin, deoxyhemoglobin, water, lipids and cytochrome oxydase can be quantified. Multiple light source and detector combinations produce images of whole tissue areas. Results of simulations and new developments in sensor design will be shown. Several clinical applications of NIRI in brain research are presented to illuminate its promising properties and to demonstrate its excellent potential in neuroscience, biomedical engineering and medicine. The seminar will be concluded by a vision of future applications and research.

The objective of epilepsy surgery is the complete resection of the epileptogenic area, i.e., the region of the cortex that has to be removed in order to make the patient seizure-free. In order to pinpoint the epileptogenic area during a presurgical evaluation, a variety of diagnostic methods is available.

The ictal EEG gives time information about the voltage distribution over the electrodes on the scalp during a seizure, and is the most commonly used method to localize the ictal onset zone. Interrater variability in the visual interpretation of ictal scalp EEG recordings is considerable, and seizure activity is often obscured by muscle and other artefacts.

We established a robust and automatic localisation method for ictal EEG activity, using the Canonical Decomposition (CANDECOMP), also known as Parallel Factor Analysis (PARAFAC), often referred to as the CP (CANDECOMP/PARAFAC) decomposition. With this decomposition, multichannel time-varying EEG is decomposed into a series of distinct 'atoms', which represent in an ideal situation distinct brain sources. We showed that one ‘atom’ can be considered as epileptic, and this ‘atom’ provides localising information.

We validated our method clinically on a set of ictal EEG recordings, and showed that the Candecomp method was more sensitive than visual interpretation of the ictal EEG’s in localising the ictal onset zone.

"Small RNAs in Arabidopsis Thaliana – a bioinformatic perspective"

Speaker: Yu Wang

Affiliation: Institute for Bioinformatics, GSF - German Research Center for Environment and Health Ingolstaedter, D-85764 Neuherberg, GERMANY

Abstract:

Small RNA (ncRNA) has recently emerged as an important landmark on the eukaryote genomic landscape. Small RNAs include microRNAs (miRNAs), small/short interfering RNAs (siRNAs) and Piwi-interacting RNAs (piRNAs). They are often less than 35 nucleotides long and play important roles in gene regulation.

In plant, large scaled deep sequencing projects have triggered a series of massive small RNA data avalanches. The number of small RNAs rose from a few thousands to hundreds of thousands. These tremendous amount data might be just a tip of a huge iceberg. David Bartel’s lab recently sequenced 340,000 small RNAs from wild type Arabidopsis thaliana plants while James Carrington’s Lab reported 200,000 sequences from both wild type plants and mutants. Surprisingly, the overlap of these small RNA data sets is very poor although each of these authors claimed that they had thoroughly sequenced small RNAs in Arabidopsis. What we can conclude from these results is that the saturation of small RNA deep sequencing has not been reached. There might be cloning bias since the labs do not use the same approaches.

A brief overview of current research on small RNAs, not limited to plants, will be presented firstly in this talk. Two topics will be discussed afterwards. The first is our work on non-conserved Arabidopsis microRNAs. We have shown that sequences similarities between members of non-conserved microRNA families are suggestive of recent evolutionary duplication events. Furthermore we found that sequence similarities are not restricted to the transcribed part but extend into the promoter regions. Sequence conservation within promoters of miRNA families as well as between miRNA and its potential progenitor gene can be exploited for understanding the regulation of microRNA genes. The second topic is our on-going research on cis-acting siRNA (casiRNA) guided DNA methylation /demethylation. We used a set of Arabidopsis thaliana silencing-deficient mutants to show that casiRNA related RNA silencing pathway negatively regulates the plant immune response. By mining small RNA databases, we identified a subset of defense-related genes that appear to be regulated by casiRNAs. The likely mechanism by which these defense-related genes are activated during the plant defense response will be presented and further integrated into a rheostatic model for epigenetic regulation of gene expression in plants.