Talks at CREATIS

Invited presentation: Ultrasound contrast agents: fundamental, imaging and new directions

Résumé :

I don't want to spoil it, but I am pretty sure there will be interesting science about nuclear imaging, nuclear detectors, image reconstruction, Monte Carlo simulation and some fun and original research about amazing physical positronium imaging. It is really a good opportunity to exchange with the leader of a very productive research group from Poland. A short bio of Prof Moskal is given below.

Jury:

Miss Magdalena Rafecas, Professor, Reviewer

Mister Jérôme Idier, DR CNRS, Reviewer

Miss Elena Resmerita , Associate Professor, Examiner

Mister Christian Morel, Professor, Examiner

Jury :

Monsieur Emmanuel BOSSY Professeur des Universités Université Grenoble Alpes Rapporteur

Monsieur Frédéric OSSANT Ingénieur de Recherche, HDR Université de Tours Rapporteur

Madame Claire PRADA Directrice de Recherche CNRS Institut Langevin (Paris) Examinatrice

Abstract:  The high frame rate vector flow imaging (HFR-VFI) has been implemented on Mindray Resona 7, which is a high-end clinical ultrasound system. Development of HFR-VFI on a clinical scanner is quite different from the theoretical researches or laboratory simulations. Image quality for both flow and gray-scale images need to be satisfied simultaneously for a general and real clinical use.

La spectroscopie ultrasonore quantitative permet de remonter à des paramètres physiques moyensdes tissus via le coefficient de rétrodiffusion. Les propriétés des diffuseurs du tissus sont obtenus parméthode inverse en utilisant des modèles de diffusion ultrasonore. Cette technique peut être utiliséepour différentes applications où l’on s’intéresse à la microstructure tissulaire, comme lacaractérisation tissulaire, ou à son évolution, comme pour le suivi d’un traitement anti-cancer.Des paramètres des diffuseurs peuvent être estimés en utilisant des modèles de diffusion.

Echographic imaging depicts anatomical structure by displaying the magnitude of the envelope-detected backscattered echoes. However, ultrasonic radiofrequency data contains a richer information content that can be exploited for constructing images of intrinsic tissue properties. In particular, spectral-based ultrasonic tissue characterization techniques allow imaging parameters such as the backscatter coefficient and the attenuation coefficient.

Given a few labelled samples, semi-supervised learning generally performs much better than supervised learning, semi-supervised learning algorithms are more robust to noise. Label Prediction via Deformed Graph Laplacian for Semi-supervised Learning is presented.  A novel curriculum learning approach, dubbed multi-modal curriculum learning, to optimize the quality of semi-supervised image classification is proposed.

NeuCube, a spiking neural network architecture, is presented for FMRI data analysis of Brain cognition, and for obstacle avoidance in prosthetic vision.

Cerebral aneurysms are dilatations of arteries feeding the brain that can present a danger of rupture, brain compression, and thrombotic occlusion of downstream vessels. A number of studies demonstrated that progression of cerebral aneurysms is affected by hemodynamic forces. Intra-aneurysmal flow dynamics can be quantified with in vivo MRI velocimetry (4D Flow MRI) or with patient-specific models based on imaging data.

Although array factorization methods have been vastly investigated as a source separation tool in disparate areas like chemometrics or remote sensing, studying multiple data set simultaneously, factorizing very large data set and identifying the extracted components solely from their spectra remain challenging tasks.

This presentation summarizes the latest scientific advances that occurred at the VITALab (Sherbrooke, Canada) and underline unsolved questions that we ought to work on in the upcoming years.  In the last two years, the VITALab (like a large proportion of the digital imaging community) has been carried by the deep learning wave.  Work has been done in various areas of video analytics as well as medical imaging.  This includes traffic analytics, motion detection, background estimation, brain tumor segmentation, and MRI cardiac analysis.