Home | Bio | Research | Software | Publications | Talks

Journal Papers

Integration and segregation of large-scale brain networks during short-term task automatization, 2016, Nature Communications. We show that short-term learning and practice causes large-scale reorganization of interactions in brain networks. DOI
Large-scale functional network reorganization in 22q11.2 deletion syndrome revealed by modularity analysis, 2016, Cortex. In patients with the 22q11 deletion syndrome, a genetic disorder, we show that the modular organisation of the visuo-spatial network and the anterior cingulate cortex is modified. This is a potentially interesting potential imaging marker for a schizophrenia risk factor. DOI
Quantitative and qualitative analysis of transient fetal compartments during prenatal human brain development, 2016, Frontier in Neuroanatomy. We show growth trajectories of fetal compartments from prenatal T1 images, and that compartments differ significantly. DOI
Correlated Gene Expression Supports Synchronous Activity in Brain Networks, 2015, Science. Using the Allen Institute human microarray database, we show that the spatial organization of functional networks correpond to correlated gene expression. We replicate in-vivo on fMRI data and SNPs, and with a model of mouse axonal connectivity. DOI. PDF: This is the author’s version of the work. It is posted here by permission of the AAAS for personal use, not for redistribution. The definitive version was published in Science 348:6240 in June 2015, DOI: 10.1126/science.1255905 Access the recommendation on F1000Prime
Multi-contrast 'connectometry': a new tool to assess cerebellum alterations in early relapsing-remitting multiple sclerosis, 2015, Human Brain Mapping. We use resting-state fMRI and Diffusion Spectrum Imaging together with two-level network analysis to show structural connectivity differences in the cerebelllum of MS patients, which are related to motor, memory, and attentional performance. DOI
Neural correlate of anterograde amnesia in Wernicke–Korsakoff syndrome, 2015, Brain Topography. We use EEG and DTI to show that Wernicke–Korsakoff syndrome is related to disconnectivity between the diencephalon and hippocampal structures. DOI
Altered cerebrovascular reactivity velocity in mild cognitive impairment and Alzheimer disease, 2015, Neurobiology of Aging. Using a hypercapnic (CO2) challenge experiment, we show that MCI and Alzheimer patients have significantly lower velocity of cerebrovascular reactivity than controls (meaning vessel contraction and dilation is slower), and that the velcocity is significantly related to cognitive performance, modulated by age. DOI
Graph analysis of functional brain networks: practical issues in translational neurosciences, 2014, Philosophical Transactions of the Royal Society B: Biological Sciences. We discuss several practical challenges and potential solutions in estimating and modelling brain networks using a graph-theoretical perspective. arXiv DOI
Identifying 22q11.2 deletion syndrome and psychosis using resting-state connectivity patterns, 2014, Brain Topography. We show that connectivity in frontal and temporal areas in particular are discriminative with high accuracy for 22q11.2 microdeletion patients, and within that group between subjects with psychotic episodes and those without. We also show how a global connectivity index relates to IQ. DOI
Head motion parameters in fMRI differ between patients with mild cognitive impairment and Alzheimer disease versus elderly control subjects, 2014, Brain Topography. We suggest that head motion in the scanner may be used as additional information regarding diagnosis. DOI
Principal components of functional connectivity: A new approach to study dynamic brain connectivity during rest, 2013, NeuroImage. We show that eigendecomposition of dynamic brain graphs can reveal temporal difference in resting-state activity between controls and multiple sclerosis patients. DOI
Machine Learning With Brain Graphs: predictive modeling approaches for functional imaging in systems neuroscience, 2013, IEEE Signal Processing Magazine. In this tutorial-style paper, we show how to do inference on graphs representing functional brain imaging data.
DOI
EEG anomalies in adult ADHD subjects performing a working memory task, 2013, Neuroscience. Using time-frequency analysis, we show that ADHD adults have clear differences from controls in event-related synchronisation when performing N-back tasks. DOI
PRoNTo: Pattern Recognition for Neuroimaging Toolbox, 2013, Neuroinformatics. We introduce the design and use of the PRoNTo toolbox for multivariate pattern analysis, including examples on several datasets. DOI
Pulsatile blood flow in human bone assessed by laser-Doppler flowmetry and the interpretation of photoplethysmographic signals, 2013, Physiological Measurement. We showed that laser-Doppler can be used to measure blood flow in bone and offered an explanation for photoplethysmographic pulsations.
Altered cortical and subcortical local coherence in obstructive sleep apnea: an fMRI study, 2012, Journal of Sleep Research. We used ReHo features with both mass-univariate testing and multivariate predictive modelling to study differences between sleep apnea patients and controls.
Classifying minimally-disabled multiple sclerosis patients from resting-state functional connectivity, 2012, NeuroImage (62). We demontrate the use of functional connectivity at rest as an imaging marker of multiple sclerosis in low-EDSS patients using our predictive modelling method. We find a significant correlation between discriminative functional connectivity alterations and white matter lesion load. DOI
Attention-related potentials allow for a highly accurate discrimination of MCI subtypes, 2012, Neurodegenerative Diseases. We show a technique based on the Attention Network Test and Independent Component Analysis that can discriminate between single-domain amnestic, single-domain dysexecutive, and multiple-domain amnestic MCI patients
Haemodynamic responses to temperature changes of human skeletal muscle studied by laser-Doppler flowmetry,2012, Physiological Measurement (33). We show that a technique to measure blood flow in muscle optically is well suited to characterise temperature regulation behaviour.
Event-related potentials and changes of brain rhythm oscillations during working memory activation in patients with first-episode psychosis, 2012, Journal of Psychiatry and Neuroscience. We show latency and band power changes in ERPs that are characteristic for first-episode psychosis patients.
Decoding Brain States from fMRI Connectivity Graphs, 2011, NeuroImage (special issue on Multivariate Decoding and Brain Reading). We propose a graph embedding approach to brain decoding, which enables connectivity-based decoding, and demonstrate the approach on a cognitive task. DOI
Bayesian networks and information theory for audio-visual perception modeling, 2010, Biological Cybernetics. We propose a graphical model structure learning approach to learn a probabilistic causal model of audio-visual perception in a localisation task. DOI
The Multi-Scenario Multi-Environment BioSecure Multimodal Database ({BMDB}), 2009, IEEE Trans. on Pattern Analysis and Machine intelligence. In this paper we describe the design, protocols, and data contents associated with the largest publicly available multimodal biometric database, collected by several European research institutions participating in the BioSecure network of excellence.
Reliability-based decision fusion in multimodal biometric verification systems, 2007, EURASIP Journal of Advances in Signal Processing. We show how bimodal fusion can be based on reliability models, and how the use of quality information helps.
Low-Level Grounding in a Multimodal Mobile Service Robot Conversational System Using Graphical Models, 2007, Intelligent Service Robotics. We apply graphical models to a multimodal (laser, video stream, speech) human-robot interaction problem.
Multimodal biometrics for identity documents (MBioID), April 2007, Forensic Science International. We describe the MBioID multimodal database, which is uniquely designed to test 6 biometric modalities (2D face, 3D face, iris, fingerprint, speech, signature) for the identity document application.
Confidence and reliability measures in speaker verification, 2006, Journal of the Franklin Institute. We compare our reliability approach to several confidence estimation methods. Our approach performs well, as it includes explicit modelling of signal quality, and more flexible distribution functions for scores and quality measures. DOI This is an invited paper

Conference Papers (peer-reviewed)

Learning with multi-site fMRI graph data, 2014, Proc. Proc. 18th Asilomar Conference on Signals, Systems, and Computers (invited paper). We propose a discriminative transform to reduce inter-scanner variability of graphs learned from fMRI data.
Recent Advances in Supervised Learning for Brain Graph Classification, 2014, Proc. IEEE Global Conference on Signal and Information Processing (GlobalSIP) (invited paper). Progress is fast in methods for comparing brain graphs. This paper covers new advances since our previous overview, "Machine Learning with Brain Graphs". DOI
Two test statistics for cross-modal graph community significance, 2013, Proc. Int Workshop on Pattern Recognition in neuroimaging. While several tests exist to determine the significance of the modularity score of a whole graph, only very few (e.g. OSLOM) can test the significance of a single community within a graph. We propose tests for the null hypothesis that any other group of the same number of vertices is as community-like (in the weak sense) in a particular graph.DOI
Functional connectivity eigennetworks reveals different temporal dynamics of multiple sclerosis patients, 2013, Proc. IEEE Int Symposium on Biomedical Imageing. We show that an eigendecomposition approach on dynamic graphs obtained from windowed correlations yields discriminative temporal information for patient versus control classification. DOI
Connectivity searchlight: a novel approach for fMRI information mapping using functional connectivity, 2013, Proc. IEEE Int Symposium on Biomedical Imageing. We propose an extension of Kriegeskorte's searchlight to take into account long-range dependencies in the data. DOI
fMRI functional connectivity estimators robust to region size bias, June 2011, IEEE Workshop on Statistical Signal Processing, Nice, France. Estimation of connectivity strength between brain regions is biased by the number of voxels. We propose corrected statistical estimators to minimise this effect. DOI
Classifying connectivity graphs using graph and vertex attributes, May 2011, Int. Workshop on Pattern Recognition in NeuroImaging, Seoul, Korea. Taking advantage of the fixed-cardinality vertex sequence property of functional connectivity graphs, we resort to a simple lossy embedding scheme based on vertex invariants to perform classification. DOI
Low-dimensional embedding of functional connectivity graphs for brain state decoding, August 2010, 1st Workshop on Brain Decoding, Istanbul, Turkey. We explore dimensionality reduction for graphs with fixed-cardinality vertex sequences, and it seems decoding accuracy can be improved. DOI
Vector space embedding of undirected graphs with fixed-cardinality vertex sequences for classification, August 2010, IAPR International Conference on Pattern Recognition, Istanbul, Turkey. We discuss limitations of graph matching algorithms for a restricted class of graphs and propose embedding techniques. DOI
Brain decoding of fMRI connectivity graphs using decision tree ensembles, April 2010, IEEE International Symposium on Biomedical Imaging, Rotterdam, the Netherlands. We propose a classification-based approach for functional connectivity graphs, using direct embedding (which works because graphs here have fixed-cardinality vertex sequences, and are low-dimensional) and a ensemble of functional trees. DOI
Impact of combining quality measures on biometric sample matching, Sept 2009, IEEE International Conference on Biometrics: Theory, Applications and Systems, Washington DC, USA. We explore the implications of low-level combination of quality measures, and conclude that using them directly as features for a classifier generally leads to superior performance.
A context-specific independence model of multisensory perception, Sept 2009, 3rd French Computational Neuroscience Conference, Bordeaux, France. We propose an elegant model of integration/segregation in multisensory perception, based on information-theoretic structure learning in directed graphical models.
Static models of derivative-coordinates phase spaces for multivariate time series classification: an application to signature verification, June 2009, IEEE International Conference on Biometrics, Alghero, Italy. We show the advantages of derivative-coordinate phase space reconstruction over lag-coordinates for time series classification problems. We recast the computation of delta coefficient in speech and signature/handwriting processing as a phase space reconstruction operation. We show results on handwritten on-line signature data.
Promoting diversity in Gaussian mixture ensembles: an application to signature verification, May 2008, First European Workshop on Biometrics and Identification Management, Roskilde, Denmark. We show techniques that can be used to optimise coverage of an ensemble of Gaussian mixture classifiers, insisting on signal processing approaches instead of traditional training set subsampling approaches such as bagging, which generally do not yield satisfactory results on GMMs.
Evaluation of speech quality measures for the purpose of speaker verification, January 2008, Odyssey 2008: The Speaker and Language Recognition Workshop, Stellenbosch, South Africa. We discuss quality measures applicable to speaker verification and show that evaluation of quality measures needs to take class-conditionality into account.
Quality measures in unimodal and multimodal biometric verification, September 2007, 15th European Signal Processing Conf. (EUSIPCO), Poznan, Poland. We propose a categorisation of quality measures and discuss their evaluation and applications in unimodal and multimodal biometric authentication. This is an invited paper.
Reliability-based voting schemes using modality-independent features in multi-classifier biometric authentication, May 2007, 7th Int. Workshop on Multiple Classifier Systems, Prague, Czech Republic. We propose the rigged majority voting scheme, which is based on training a meta-classifier (for instance a reliability model) and letting the meta-classifier vote instead of the base classifier. By doing so, we can improve both the lower and upper bounds on accuracy of majority vote. This can be attributed to the use of quality measures in the meta-classifier.
Reliability estimation for multimodal error prediction and fusion., June 2007, 7th Int. Workshop on Pattern Recognition in Information Systems (PRIS 2007), Funchal, Portugual. We show that reliability estimation can be used to effectively take advantage of the rejection-error tradeoff in multimodal fusion.
Quality dependent fusion of intramodal and multimodal biometric experts, April 2007, SPIE Defense and Security Symposium, Orlando, USA. We show that using signal quality information (modality-specific quality measures) is beneficial for fusion.
Speaker Verification with Confidence and Reliability Measures, May 2006, IEEE International Conference on Speech, Acoustics and Signal Processing, Toulouse, France. We show how our proposed model of decision reliability yields error prediction superior to that of classical confidence measures, through the use of quality measures and flexible distribution parameters.
Writer Identification for Smart Meeting Room Systems, February 2006, 7th IAPR Workshop on Document Analysis Systems, Nelson, New Zealand. Features specific to writer identification, proposed by the University of Bern, improve performance over our state-of-the-art GMM-based signature verification classifier.
A multimodal biometric authentication framework, October 2005, 3rd COST 275 Workshop, Hatfield, UK, in which we give details on our Java-based distributed biometric user authentication framework.
Local and Global Feature Selection for On-line Signature Verification, August-September 2005, 8th IAPR International Conference on Document Analysis and Recognition, Seoul, Korea, in which we extend our previous work on global feature selection to local features and find that they, too, can be selected appropriately by our algorithm. DOI
Error Handling in Multimodal Biometric Systems using Reliability Measures, September 2005, 13th European Signal Processing conference, Antalya, Turkey, in which we use reliability measures to break ties between face and speech modalities in a multimodal identity verification system. This paper was selected as one of 10 finalists in the student paper competition.
Global feature selection for on-line signature verification, June 2005, 12th International Graphonomics Society conference, Salerno, Italy, in which an algorithm based on sequential forward floating search and a modified Fisher Ratio is proposed to perform feature selection on global on-line signature features. This paper was awarded a graduate student award for best graduate student presentation from the International Society of Motor Control.
Graphical models for dialogue repair in multimodal interaction with service robots, May 2005, 8th COST276 Worshop, Trondheim, Norway, in which we use Clark and Schaefer's grounding model of conversation to develop a multimodal Bayesian network fusing video, speech, and laser data. The goal is to make the robot aware of communication failures, and to proceed with appropriate repair depending on the grounding state of the conversation. Different repair strategies are encoded using decision networks.
A probabilistic measure of modality reliability in speaker verification, March 2005, IEEE International Conference on Speech, Acoustics and Signal Processing, Philadelphia, USA, in which a graphical model (Bayesian network) is used to learn the relationships between classifier errors, score distributions, and acoustic conditions, quantify the reliability of the speech modality, and subsequently used to obtain more robustness to noise in a speaker verification task. This paper won a best student paper award in the speech processing category. DOI
On-line signature verification resilience to packet loss over IP networks, March 2004, COST275 Workshop on Biometrics on the Internet: Fundamentals, Advances and Applications, Vigo, Spain, in which strategies for coping with packet loss for distributed biometric systems are explored.
Gaussian Mixture Models for On-line Signature Verification, Nov. 2003, Proc. ACM int'l conference on Multimedia 2003, Workshop on Biometric Methods and Applications, USA, Berkeley, in which it is shown that GMMs are a promising model for on-line signature verification. DOI

Abstracts presented at conferences (non peer-reviewed)

Gene expression correlations map to functional brain networks, 2016, Abstract at the 89th Annual Meeting of the German Neurological Society (invited talk)
Graph-based statistical testing for imaging genomics of functional networks, 2016, Abstract at 12th Annual International Imaging Genetics Conference (invited talk)
Graph analysis of the motor network after primary motor cortex stroke: longitudinal changes in node strengths and recovery, 2016, Poster at the 2​nd European Stroke Organisation Conference
Hemispheric asymmetry in the processing of left/right rotations during spatial updating of scenes, 2016, Abstract at the Organisation for Human Brain Mapping Annual Meeting
Memory task activation and resting-state connectivity are correlated across normal aging and MCI, 2016, Abstract at the Organisation for Human Brain Mapping Annual Meeting
Aberrant resting-state functional connectivity in functional neurological disorder, 2016, Abstract at the Organisation for Human Brain Mapping Annual Meeting
Sequential mapping of the motor network after primary motor cortex lesion: changes in node strengths and recovery, 2016, Abstract at the Organisation for Human Brain Mapping Annual Meeting
Regularizing a linear SVM to minimize scanner influence in multi-centric neuroimaging data, 2015, Poster at the 5th Int. Workshop on Pattern Recognition in NeuroImaging
Predicting pure amnestic mild cognitive impairment conversion to Alzheimer's disease using joint modeling of imaging and clinical data, 2015, Abstract at the 2015 International Conference on Brain Informatics and Health
Genetic bases of functional brain networks, 2015, Abstract at the 10th Alpine Brain Imaging Meeting (selected for oral presentation)
Memory processing-related functional networks relevant to mild cognitive impairment, 2015, Abstract at the 10th Alpine Brain Imaging Meeting
Thalamic neurodegeneration disconnects resting-state networks in early multiple sclerosis: results from a data-driven multimodal approach, 2014, Abstract at the Alpine Brain Imaging Meeting (ABIM) 2014
Identifying amnestic mild cognitive impairment by activation patterns during an associative memory task, 2014, Abstract at Journees de Printemps de la Societe de Neuropsychologie de Langue Francaise (oral presentation)
Brain atlas choice impacts classification of amnestic mild cognitive impairment patients using resting-state functional connectivity, 2014, Abstract at the Resting States and State Dependent Information Processing in Health and Disease conference
Combined cerebrovascular reserve and functional connectivity MRI analysis to discriminate Alzheimer Disease, mild cognitive impairment and healthy controls, 2011, Poster at the annual joint meeting of the Swiss Society of Neuroradiology and Swiss Society for Behavioural Neurology
A "guided searchlight" approach in classifying amnestic mild cognitive impairment patients, 2014, Poster at the Organisation for Human Brain Mapping annual meeting (OHBM)
Local topological properties of brain connectivity in minimally-disabled multiple sclerosis patients, 2014, Abstract at the Annual Meeting of the Organization for Human Brain Mapping (OHBM)
Cerebellar connectomics provide new biomarkers in early multiple sclerosis, 2014, Abstract at the International Society for Magnetic Resonance in Medicine annual meeting
Structural and functional alterations in presymptomatic Alzheimer's disease revealed by multivariate pattern analysis, 2013, Abstract at the nanoymposium on Alzheimer's Disease: Imaging and Biomarkers (selected for oral presentation), Annual meeting of the Swiss Society for Neuroscience
Principal component analysis identifies building blocks of dynamic brain connectivity during rest, 2013, Poster at the Annual Meeting of the Organization for Human Brain Mapping
A novel approach to study dynamic networks in the healthy and diseased human brain, 2013, Abstract at 4th Workshop on Complex Networks (CompleNet 2013)
Coordination of the default-mode network is fostered by neurotransmitter-related genes, 2013, Poster at the annual conference of the Stanford Institute for Neuro-Innovation and Translational Neurosciences
Multi-class fMRI decoding: generalised information mapping and multivariate thresholding, 2013, Poster at the Annual Meeting of the Organization for Human Brain Mapping
Multivariate pattern interpretation using PRoNTo, 2013, Poster at the Annual Meeting of the Organization for Human Brain Mapping
PRoNTo: Pattern Recognition for Neuroimaging Toolbox, 2013, Abstract at NIPS 2013 Workshop on Machine Learning Open Source Software (MLOSS)
Pattern Recognition for Neuroimaging Toolbox, 2013, Abstract at International Workshop on Advances in Regularization, Optimization, Kernel Methods and Support Vector Machines:
theory and applications (ROKS)
Pattern Recognition for Neuroimaging Toolbox, 2013, Video Journal of Machine Learning Abstracts
fMRI scan-by-scan decoding of cognitive states using functional connectivity graph wavelets, 2012, Abstract at Annual Meeting of the Organisation for Human Brain Mapping
Scan-by-Scan Decoding of Cognitive States From fMRI Data Using Whole-Brain Functional Connectivity, 2012, Poster at the Alpine Brain Imaging Meeting (ABIM)
Eigennetworks as signatures of brain connectivity dynamics, 2012, FENS-IBRO-HERTIE Winter School on Brain Dynamics and Dynamics of Brain Disease
Functional Disconnectivity at rest in 22q11.2 Deletion Syndrome, 2012, Poster at Lemanic Neuroscience Annual Meeting
Functional Disconnectivity at rest in 22q11.2 Deletion Syndrome, 2012, Poster at Annual Scientific Meeting of the Swiss national MD-PhD program (prize for best poster)
Practical issues in functional connectivity analysis: advice for practitioners, 2011, Poster at the 5th Alpine Brain Imaging Meeting
Brain decoding using fMRI connectivity analysis: theory and applications, 2010, Poster at the 4th Alpine Brain Imaging Meeting
Inter-subject brain state decoding by classification of fMRI functional connectivity graphs, 2010, Poster at the Annual Meeting of the Swiss Society for Neuroscience (won the Volker Henn best poster award)
Classification of functional connectivity graphs for multi-subject brain state decoding, 2010, Poster at the Annual Meeting of the Organization for Human Brain Mapping
Processing fMRI functional connectivity data: brain network decoding and analysis, 2010, Poster at the Lemanic Neuroscience Annual Meeting
Listen to the brain: decoding of distributed fMRI patterns in human auditory cortex, 2010, Poster at the Annual Meeting of the Organization for Human Brain Mapping
Graphomotor pattern recognition in a memory clinic setting, Oct. 2009, Swiss Society of Gerontology annual meeting.
Classifying multivariate time series with static Bayesian networks and delta features: a phase space perspective, 2008, Poster at the IM2 National Center of Competence in Research Summer Institute
Bayesian networks for multiple classifier biometric verification systems, 2007, Poster at the Winter Institute of the National Center of Competence in Research "Interactive Multimodal Information Management" (IM2)
Biometric data quality and expert confidence dependent fusion of multiple modalities, 2006, Oral presenation at the Biosecure Vigo Workshop
Estimating decision reliability in biometric user verification, 2005, Oral presentation and Poster at the Summer Institute of the National Center of Competence in Research "Interactive Multimodal Information Management" (IM2)