Vous êtes ici : Accueil » Kiosque » Annonce


Mot de passe : 

Mot de passe oublié ?
Détails d'identification oubliés ?


10 janvier 2017

Stage de Master : Atrial fibrillation analysis via tensor decompositions

Catégorie : Stagiaire

MSc Internship Proposal 2016-2017

“Atrial fibrillation analysis via tensor decompositions”


Context: The internship will take place at the I3S Laboratory, in the context of an ongoing interdisciplinary collaboration with the Cardiology Department of Monaco “Princess Grace” Hospital.

Topic description: Atrial fibrillation (AF) is the most common sustained arrhythmia encountered in clinical practice, especially affecting the elderly. Held responsible of up to 25% of strokes, this cardiac condition is considered as the last great frontier of cardiac electrophysiology as it continues to puzzle cardiologists. Previous research has shown that the noninvasive analysis of AF is facilitated by first arranging the observed surface electrocardiogram (ECG) recordings in the form of two-dimensional matrix structures, with rows and columns typically representing the spatial and temporal dimensions. Such matrices are then factorized into underlying source signal contributions using suitable matrix decomposition techniques such as principal component analysis or independent component analysis. This matrix decomposition approach has proven useful for a variety of purposes including artifact cancellation, AF complexity assessment and therapy outcome prediction.

A recent line of research aims to take a step forward in this multidimensional approach to cardiac signal analysis by considering data structures with more than two dimensions, known as multi-way arrays or tensors. As compared to matrix techniques, tensor decompositions present some remarkable features such as essential uniqueness and rank possibly exceeding the data dimensions. Preliminary results in the context of noninvasive atrial signal extraction are promising [1, 2, 3], but require an accurate selection of the model parameters defining the tensor model.

The MSc internship will deepen this novel characterization by searching for automatic parameter selection schemes for the tensor model in this biomedical application. The proposed solutions will be tested using realistic synthetic data, and then validated on actual recordings from an AF ECG database acquired at several stages of ablation procedures. Such recordings will allow us to test the robustness of the developed algorithm in a variety of real clinical scenarios.

If successful, the internship could be prolonged into a PhD thesis.

Pre-requisites: Candidates will require a strong background in mathematics (especially matrix algebra), statistics and signal processing. Previous experience with biomedical signals, in particular the ECG, and familiarity with cardiac electrophysiology will also be desirable.

Contact: Vicente Zarzoso, I3S Laboratory, University of Nice Sophia Antipolis (zarzoso@i3s.unice.fr).


[1] L. N. Ribeiro, A. R. Hidalgo-Muñoz, V. Zarzoso, “Atrial signal extraction in atrial fibrillation electrocardiograms using a tensor decomposition approach”, in: Proc. EMBC-2015, 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Milan, Italy, Aug. 25-29, 2015, pp. 6987-6990.

[2] L. N. Ribeiro, A. R. Hidalgo-Muñoz, G. Favier, J. C. M. Mota, A. L. F. de Almeida, V. Zarzoso, “A tensor decomposition approach to noninvasive atrial activity extraction in atrial fibrillation ECG”, in: Proc. EUSIPCO-2015, XXIII European Signal Processing Conference, Nice, France, Aug. 31-Sept. 4, 2015, pp. 2576-2580.

[3] L. N. Ribeiro, A. L. F. de Almeida, V. Zarzoso, “Enhanced block term decomposition for atrial activity extraction in atrial fibrillation ECG”, in: Proc. SAM-2016, 9th IEEE Sensor Array and Multichannel Signal Processing Workshop, Rio de Janeiro, Brazil, July 10-13, 2016.


Dans cette rubrique

(c) GdR 720 ISIS - CNRS - 2011-2015.