Innovative Projects Realized

Brain Connectivity Analysis Using Machine Learning

Project description:

The selected interns will work with graduate students on developing novel methods to analyze functional and structural brain connectivity. Two applications are considered for this project: brain fiber clustering and dynamic functional connectivity analysis.

For the first application, recent machine learning techniques, based on dictionary learning, will be used to group the white matter fiber tracts into prominent bundles. Extracting these bundles is essential to perform a high-level analysis of structural brain connectivity. The methods developed for this application will be tested on real diffusion MRI data. The role of the intern for this application will be to assist the graduate students in programming and testing core functions of the proposed method. The intern will also work on developing programs to visualize the clustering results.

The second application targeted by the project is related to multi-subject dynamic functional connectivity (FC). So far, functional connectivity analyses in resting-state fMRI have been dominated by static FC, which compute the correlation between whole time-series from different voxels. Recently, dynamic FC for resting-state fMRI has been proposed, by considering small spatio-temporal windows from individual fMRI data and computing how the correlation changes between the signals captured in these windows. The project will use co-clustering techniques to jointly analyze the data from multiple subjects. Once again, the intern will be involved in the implementation and testing of the proposed method.

Analysis and implementation of behavioral models for the development of advanced GaAs HBT/HEMT RFIC Power Amplifiers

RFIC PA (RF Integrated Circuit Power Amplifier) designers often face the difficult challenge of bridging the gap between simulation data and experimental results, which negatively impacts the design convergence and time-to-market considerations during the development of complex amplifier structures. This is particularly severe when dealing with nonlinear circuit techniques during the practical development phase. Design oriented behavioral models (e.g. [1], [2]) that allow bridging this gap in a time efficient way are valuable to RFIC PA designers. This project aims at developing a design oriented behavioral model that is suited for the development of standalone RFIC PA’s employing complex architectures, in GaAs HBT and other technologies.

The project integrates the framework of on-going investigations on complex GaAs/InGaP HBT/BiHEMT RFIC PA structures for wireless communications, in collaboration with a worldwide leading supplier of RFIC PAs. A behavioral model that suits the task of correlating experimental data and simulation results during the practical phase of RFIC PA development will be analyzed, improved and implemented within design platforms (Agilent – ADS Simulation software and other experimental development tools). A methodology that accompanies the behavioral model will be proposed and validated in a way that demonstrates the impact on the design convergence during RFIC PA development.

Starting from a well-defined RFIC PA architecture, RF electrical performance objectives (e.g. gain flatness, power added efficiency, linearity and stability), a preliminary feasibility study, an existing basic behavioral model (e.g. [1], [2]) and mathematical formulations, new formulations (or adapted from formulations proposed in other work) that describe the nonlinear behavior of RFIC PA’s will be derived and an improved behavioral model will be developed. On the one hand, the model will aim at enhancing the simulation capabilities with ADS by incorporating experimental RF transistor device and circuit blocks characterization data, as well as PA system experimental results obtained from actual PA implementations. On the other hand, the same model will be adapted to allow efficient diagnosis and troubleshooting of PA performances by the designer during the development phase, with the aim of speeding the design convergence.

[1] S. Sharma, N.G. Constantin, “Formulations for the Estimation of IMD Levels in an Envelope Feedback RFIC Amplifier : An Extension to Dynamic AM and PM Behavior,” IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, Nov. 2013.

[2] N. G. Constantin, K. H. Kwok, H. Shao, C. Cismaru, and P. J. Zampardi, “Formulations and a Computer-Aided Test Method for the Estimation of IMD Levels in an Envelope Feedback RFIC Power Amplifier,” IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, Dec. 2012.

Tongue motion analysis using ultrasound images

Ultrasound imaging is an ideal tool for studying the tongue motions involved in speech as it provides very rich information while being relatively inexpensive and non invasive. Its use has thus become widespread among speech scientists.

The analysis of ultrasound video sequences of the moving tongue poses a number of challenges pertaining to the reliable extraction, tracking and analysis of the moving tongue contour. The goal of this project is to develop new automated and interactive statistical tools to analyze tongue shape and motion as they pertain to interesting questions in the field of speech science. An example of such a question might be, “how does tongue motion differ in children and adults during speech?”. To answer such questions, the student will study, implement and compare various techniques from the field of pattern recognition.

The project will be carried out in collaboration with speech scientist Prof. Lucie Ménard, who leads the Phonetics Laboratory at University of Quebec in Montreal’s department of linguistics.

Real-Time multi-view rendering from 3D stereoscopic video

With the steady production of 3D movies, there is an increased availability of 3D stereoscopic video content, paving the way for the deployment of 3DTV to the general consumer market. Auto-stereoscopic displays, those that produce 3D without the need to wear glasses, are a viable technology for the home environment and are obtaining a lot of attention from the media production and consumer electronics industries. One of the main issues to solve for the introduction of auto-stereoscopic 3DTVs in the home market is the lack of content specially prepared for this kind of display. Auto-stereoscopic displays require several views, more than two and normally around nine, to present multiple points of view to the viewer, generating a more realistic and comfortable 3D representation of the scene. Given that most of the content currently produced is based on a two-view format (S3D), there is an important need for high quality algorithms for converting two-views stereoscopic video content to multi-view format for auto-stereoscopic displays. The additional requirement of real-time conversion for the broadcasting industry adds a layer of complexity to the conversion problem.

In this project we aim at developing a set of tools to perform the conversion of video content from two-view to multi-view format in real-time. The intended application is the broadcasting industry. The amount of data to process, HD and UHD (4K), and the high image and depth quality required for broadcasting applications makes this problem a difficult one. The conversion process includes three main steps: recovering the depth information from the stereoscopic video source, rendering of the additional views required for auto-stereoscopic displays based on the recovered depth and the viewing conditions and post-processing the images for presentation in the auto-stereoscopic display. Each of these steps requires the application of a number of image and video processing algorithms to obtain the desired results.

This particular project will focus on the join extraction of the depth information and rendering of the new views in real-time. The extraction of depth information from stereoscopic video sources is a well studied problem normally referred to as the stereo matching or stereo correspondence problem. There is, however, not a single algorithm that performs well for all types of videos. Another problem to tackle is the complexity of the available algorithms. Most of the proposed algorithms are highly complex and are not suitable for real-time implementation. The combination of the depth extraction step with the view rendering process is a promising avenue to improve both, the quality of the reconstructed images and the processing time of the whole process.

Another aspect that will be considered in the project is the choice of depth parameters to ensure the comfort of the viewer. Viewer comfort for stereoscopic imaging is one of the most important research topics in the 3DTV research community, and ensuring that the produced depth is consistent with the original scene and comfortable to view is one important requirement of the project.

Development of a simulation platform for soft and deformable human tissues

Introduction:
Medical simulation is a rapidly growing field, having a positive impact on how healthcare providers are being trained and evaluated. Over the last ten years, there has been a 15-fold increase in the number of medical simulation centres (1400 worldwide today). Surgical simulation allows training outside the operating room (OR), thereby minimizing patient risk, facilitating adoption of new surgical techniques and assuring efficient OR use, for improved patient care. Current surgical simulators employ first-generation virtual reality (VR) technology involving the manipulation of simplistic anatomical structures with virtual surgical instruments. For surgical specialties requiring high precision and acute perception, such as ENT (ear-nose-throat) virtual simulation requires more realistic representation of both anatomical structures and surgical procedures with a diversity of surgical tools.

In ENT, high precision and acute perception are necessary to avoid damage to cerebral tissues and vital structures thereby minimizing the risk of debilitating conditions and potentially fatal outcomes. VR simulation is widely recognized as a valuable tool for reducing adverse effects from surgery, by allowing for training and rehearsal outside the OR, without consequences for the patient.

Project:
In this project, we aim to develop the foundations of a physics-based VR surgical simulation engine focused on the task of soft and deformable tissues resection and removal. Physics-based simulation, such as finite elements, is especially important for high precision and acute perception surgeries such as ENT. The approach provides higher realism for visual and touch simulation of the physical situation, by relying on continuum mechanics fundamentals. The simulation is achieved through real-time resolution of the corresponding mathematical equations.

In the project, we’ll need to develop software that comprises real-time finite element computation integrating tissue biomechanical models and anatomical models derived from patient images. In addition, the simulation software engine should be able to interactively accomplish tasks in surgical scenarios using a variety of surgical tools. Implementation of new surgical tools requires development of scenario data as well as software development in the core engine, both of which are specific to the surgical tool.

The simulation software engine would use finite elements to calculate tool-tissue interactions, currently handling real-time simulation of an operating field of 3 mm resolution and 5,000 finite elements (brain tumour resection on a stand-alone computer). The engine should simulate tissue deformation, removal and bleeding and includes basic tissue manipulation and hyper-elastic tissue deformation models, derived from biomechanical data obtained on animal, human and virtual tissues.

Characterization of braided composite aircraft structural components

With the increasing usage of composite in civil aviation, the industry is looking for affordable manufacturing techniques to produce high quality structural components. To achieve this goal, one approach is to use automatized textile techniques to quickly assemble fibre architectures tailored to the application. In this project, braiding is used to construct fuselage frame fibre preforms with variable geometry. The student will participate in the production of the braids using a 144 carrier braiding machine and a robotized gantry system at the Saint-Hyacinthe (Qc) laboratory. The student will also help characterize the meso-structure of the braid (Montreal Lab) in order to guide the design of optimal structures.

Development of a Sensor Network for Pedestrian Data Collection

In this ongoing research project we are interested in collecting data at the pedestrian facility level (e.g. train station, outdoor festival area) using automated techniques. This means that the data gathering process should be able to know where a particular pedestrian entered, what activities they performed, and where they exited the system. We need a system of sensors that can cover the entire study area and are able to communicate with each other and the central mobile server.

We are developing a communication network architecture that can achieve this purpose. The deployed sensors will be part of a IEEE 802.11 (also know as WiFi) communication network that will use Internet Protocol (IP) layer for network identification and connection, while User Datagram Protocol (UDP) layer for the data communication. To ensure security of the network, the UDP packets will be encrypted using Datagram Transport Layer Security (DTLS) protocol outlined in RFP43471. Here we will take advantage of the initial work done on the mobility over TCP/IP networks by myself.

The sensors may include, cameras, depth sensors, infra-red, thermal and other sensors. Each of them will have a raspberry pi board attached, which will act as a controller of the sensor and as a wireless communication point. This board will also do some level of the onsite processing of the data.

The network will be able to communicate two-ways between the sensors and with the mobile server. This will ensure that most of the pedestrian identification and tracking can be performed at sensors in a cooperative manner. The pedestrian traces will then be gathered and fused at the mobile server so to eventually be transmitted to a cloud based data warehouse.

Experimental validation of a novel medical device tracking technology

Hepatocellular carcinoma (HCC) is the 5th most common cancer worldwide with 500,000 new cases per year and the highest mortality rate (>97% in 5 yrs). Intravenous chemotherapy is limited and clinical outcomes are generally poor with a median survival rate of less than one year. Transcatheter arterial chemoembolization (TACE) is an image guided interventional oncology procedure that is the mainstay of intermediate stage HCC therapy. It has been shown to control symptoms, however the rapid distribution of the drug in the whole body prevents high intra-tumoral drug concentrations to be sustained. Still, targeting tumor cells by carrying a specific endovascular drug or radioisotopes delivery at the site of the HCC tumor mass is a challenging problem due to the complexity of the liver’s vasculature and current limitations to contrast injections which is unsuitable for some patients.

To deliver the appropriate amount of chemoembolizing agents, groundbreaking work in therapeutic drug delivery and minimally invasive interventions must find the appropriate tracking technologies to facilitate translation in the interventional workflow, with automated image-based multimodal fusion and intra-arterial catheter localization. To better assist interventional radiologists follow the catheter’s position and reach the tumor site with increased accuracy and higher confidence, our ongoing research project is developing an innovative solution for intra-operative medical instrument guidance and real-time non-rigid registration for image-guided interventional procedures. The envisioned platform includes a device composed of optical fibre Bragg gratings with a unique helicoidal core to infer in a real-time fashion, the 3D shape of a catheter inside the body to be visualized by the interventionist. This minimally invasive technology respects a number of critical elements such as offering computational accuracy, integration in complex environments in the operating room or X-ray interventional suite, and real-time interactions for timely information feedback. Due to the stringent accuracy constraints to preserve the alignment with the virtual pathways, real-time motion compensation is a key focus of our efforts. Despite remaining sensitive to external fluctuations, the shape sensing optical fibre based on multiple fibre gratings detects reflected light signals which are integrated into the catheter in order to monitor the dynamic 3D shape of the vessel influenced by breathing or compensate for involuntary movement.

This research project provides a new pathway in device localization, motion estimation and multimodal image fusion during guidance procedures, thereby simplifying the surgical environment and clinical workflow by introducing automated workflows. This will provide a unique opportunity to improve the robustness and reproducibility for introducing flexible devices in vessels and coronaries, and guiding them to a pre-identified target.

3D printing of polymer-based microelectromechanical systems

There is currently a big craze in 3D printing, a transformative technology that will most likely change the way we fabricate a plethora of products, ranging from toys to complex aircraft components. The research project addresses the entire span of high-end printing applications, ranging from material design to the fabrication process. On the material side, the intern will mix the best nanoscopic materials inside different kinds of plastics for improved electrical and mechanical properties while innovating in 3D printing methods using commercial and custom-made printers. The student might also be involved in the design and fabrication of printing accessories.

A promising qubit for quantum computation: excitons bound to small molecules embedded in semiconductors.

Single impurity atoms in semiconductor crystals can be spatially resolved and studied individually. Carefully selecting the nature of the impurity and the host material, an impurity center composed of one, two or three atoms can bind electrons and holes, thereby forming an exciton bound to a quantum structure. Although the electronic properties of these atomic-size quantum dots are similar to those of conventional quantum dots composed of tens of thousand atoms, their size is comparable to the volume of a few atoms. This offers excellent opportunities for the realization of a spin-based qubit of atomic dimensions for the field of quantum computation.

Using ultrafast laser pulses and optical spectroscopy techniques, we have recently demonstrated that it was possible to initialize an exciton qubit and manipulate its state over the whole Block sphere. Doing so revealed a very high optical dipole moment and a very low power induced dephasing, making this system a very attractive building block for high-fidelity quantum operations.

Optical transitions and selection rules in 2D semiconductors. (New)

Inspired by the unusual properties of graphene, the search for other 2D materials has unveiled that MoS2 monolayers also offer unusual and spectacular characteristics. In contrast to gapless graphene, MoS2 is perfectly adapted for logic electronics and optoelectronics with a gap of 1.8 eV. In addition, excellent transistor performance, long device lifetime, and high mechanical strengthmake MoS2 an ideal material for the development of transparent displays and other flexible electronic applications. However, the properties of this promising 2D material remains to be unveiled, as little is known on its electrical, optical and mechanical properties. One of the most fundamentally characteristic of a material is its band structure close to the Fermi level: it determines most of the optical and electronic properties.

This project consist in providing the experimental values for reconstructing the band structure. Using electroreflectance spectroscopy in a wide spectral range, the student will determine all optically allowed optical transitions, along with their polarization selection rules. Using this information, the student will unveil the evolution of main critical points as a function of the number of monolayers and as a function of perturbations like temperature, strain and magnetic field.

Modeling the thermodynamic and physical properties of ternary mixtures of ionic liquids (i.e. room temperature molten salts)

Modeling the thermodynamic properties (including phase equilibria) and physical properties of multicomponent systems is of great industrial importance nowadays. The metallurgical and chemical industries commonly use thermochemical packages to simulate chemical reactions and phase equilibria for process development. The thermodynamic and phase equilibrium modeling of multicomponent salt systems consists in expanding the Gibbs energy of a phase as a function of temperature, pressure and composition, and finding a global energy minimum. Model parameters of the Gibbs energy functions are obtained by optimizing their values in order to best reproduce simultaneously the experimental data found in the literature (enthalpy, liquidus, etc.). Parameters form a database and the models are used to predict phase equilibria and thermodynamic properties in the multicomponent system. Models for the density, viscosity and electrical conductivity of multicomponent inorganic molten salts, all linked to the thermodynamic model that gives an estimation of the structure of the melt, were previously developed and applied successfully to inorganic electrolytes such as NaCl-KCl-MgCl2-CaCl2 and NaF-AlF3-CaF2-Al2O3-LiF-MgF2.
“Room temperature molten salts” (or “ionic liquids”) are involved in numerous potential applications (solvents and catalysts, electrochemistry,…). They are usually composed of a large organic cation and an inorganic anion, and they melt at relatively low temperatures (often below room temperature). Since 2000 the research on “ionic liquids” has grown dramatically and it has focused on the design of single compounds with “tailor-made” properties. As suggested by Plechkova and Seddon, “ionic liquid” ternary mixtures may be considered, where the 1st component would control and define the chemistry of the system, the 2nd component would allow fine tuning of the physical properties (such as density and viscosity) of the system, and the 3rd component would be cheap and inert, thus lowering the global cost of the system. “Ionic liquid” mixtures have been studied relatively little, and most of the existing studies correspond to binary mixtures.
The proposed project consists in developing models and databases of parameters for the prediction of phase equilibria (in particular, the liquidus temperature) and physical properties (mainly density and viscosity) of ternary mixtures of the type CX-CY-CZ (where C is a large organic cation such as 1-alkyl-3-methyl-imidazolium, and X, Y and Z are three small anions such as Cl-, NO3- and CH3SO3-). The thermodynamic properties (mainly phase diagram) of the liquid solution will be modeled with the Modified Quasichemical Model in the Pair Approximation. The experimental data necessary for the calibration of the model will be provided by QUILL (Queen’s University Ionic Liquid Laboratories), located in Northern Ireland. Depending on the progress in this project, the density and viscosity of the “ionic liquid” ternary mixtures will also be modeled.