Team(s) Photo September 2020, amidst COVID-19 Working-from-home
Kâmil Uludağ, Ph.D.Principal Investigator
Kâmil Uludağ studied from 1992 till 1997 Physics at the Technical University of Berlin. He completed his Ph.D. in Physics in 2003 on Near-Infrared Optical Spectroscopy (Humboldt University, Berlin) and moved for a postdoc position to the Center for Functional MRI (UCSD, San Diego, USA) to work on the physiological and physical basis of functional MRI. In 2004, he was appointed Head of Human Brain Imaging group at the Max-Planck-Institute for Biological Cybernetics, Tübingen, Germany. From June 2010 to December 2018, he was Associate Professor in the Faculty of Psychology & Neuroscience and Head of the Department of Cognitive Neuroscience continuing his work on the fundamentals of fMRI utilizing the new Ultra-High Field human MRI scanners (7 and 9.4 Tesla). Since May 2019, he is Full Professor at the Department of Medical Biophysics, University of Toronto.
Dr. Uludağ is on the editorial board of five neuroimaging journals, served from 2011 to 2013 as Annual Meeting Committee Member of the International Society for Magnetic Resonance Imaging in Medicine (ISMRM) and was elected Chair of the Current Issues of Brain Function study group. He recently edited a textbook “Functional MRI: from Nuclear Spins to Brain Functions” (Publisher: Springer).
Dr. Uludağ’s laboratory combines artificial intelligence approaches with 1.5 and 3T MRI big data in order to answer fundamental neuroscience questions and to develop biomarkers for clinical applications. Furthermore, he uses deep learning methods and generative models to improve the effectivity of MR image acquisition and reconstruction, promising to advance our understanding of the physical and physiological basis of MRI. As Co-Director of the Slaight Family Centre for Advanced MRI at the Toronto Western Hospital, Dr. Uludağ advises the clinical research groups in their studies of the brain and spine.
Annie is an administrative assistant at the BRAIN-To lab since 2019. Prior to this, she worked at the STTARR Innovation Centre and Cyclotron & Radiochemistry Facility at UHN as an administrative assistant for 8 years. Annie graduated in Medical Office Administration with Honours from Centennial College.
Lars Kasper, Ph.D.Scientific Associate Personal Website
Dr. Lars Kasper is a scientific associate at the BRAIN-To lab since 2020. He has more than 12 years of expertise in technological development and neuroscientific application of structural and functional MRI in humans, at both high (3 Tesla) and ultra-high (7 Tesla) magnetic field strength.
Lars graduated with a MSc in Physics (German "Diplom-Physiker") from the University of Goettingen, Germany, in 2008. His thesis at the Max Planck Institute for Biophysical Chemistry (Prof. Jens Frahm, Prof. Susann Boretius) explored MRI relaxometry for ultra-high field animal imaging (9.4T).
His PhD (2008-2013) at the Institute for Biomedical Engineering (ETH Zurich and University of Zurich, Switzerland) comprised MRI research at the interface between technological development and neuroscientific application, conducted in the Translational Neuromodeling Unit (Prof. Klaas Enno Stephan) and MR Technology and Methods Group (Prof. Klaas P. Pruessmann). His PhD thesis on "Noise Reduction in Functional MRI utilizing Concurrent Magnetic Field Monitoring" covers MR sequence development and image reconstruction aspects, as well as physiological noise modeling and preprocessing considerations. A key insight is the integrated perspective on acquisition and post-processing and analysis methodology, enabled by the advent of NMR probe-based magnetic field monitoring as novel technology.
His post-doctoral (2014-2016) as well as senior research fellow (2016-2019) appointments at the Institute for Biomedical Engineering (ETH Zurich and University of Zurich, Switzerland) deepened this translational approach and converged in ultra-high resolution spiral imaging applications for anatomical MRI and layer-specific functional MRI.
Improving MRI by Mechanistic Models of Signal Encoding
Magnetic resonance imaging (MRI) offers a unique opportunity to non-invasively study the human brain, in a favourable window of spatiotemporal resolution. But to expand its clinical and neuroscientific scope, MRI has to become faster and more robust. A key ingredient to this end is a mechanistic understanding of the imaging physics, and its consideration in the way we acquire and reconstruct MR images. This comprises
- the choice of optimal sequence and sampling strategy for the target application (e.g., spiral trajectories for short echo-time)
- a detailed characterization of both hardware imperfections (e.g., heating/field drift and trajectory errors) and subject-related noise sources (e.g., breathing, motion) giving rise to image artifacts,
- the accurate inclusion of this information into an expanded signal model of the MR imaging process and
- advanced image reconstruction techniques capable of inverting these intricate models.
This integrated approach leads to better imaging data tailored to the analysis and diagnostic purpose of the MR exams. The characterization of each scan's imaging peculiarities allows for better comparability of multi-site studies in the era of big data. At the other end of the spectrum, personalized medicine could require individualized MR imaging protocols that are enabled by flexible and extendable signal models and reconstructions.
Zhe "Tim" Wu, Ph.D.Postdoctoral Fellow
Zhe Wu (Tim) is a Postdoctoral Fellow at Techna Institute of the University Health Network (UHN). His Ph.D. project was on fast MR quantitative myelin water imaging. In 2018, after obtaining his Ph.D. degree of Biomedical Engineering from Zhejiang University in China, he worked as a Postdoctoral Researcher at the Montreal Neurological Institute and Hospital (MNI, or the Neuro) of McGill University in Montréal, Canada. In June 2019, he moved to the UHN and joined Dr. Kâmil Uludağ’s group. Dr. Wu is a specialist in MR sequence programming and image reconstruction, with experience on both GE EPIC system and Siemens IDEA platform. Dr. Wu has published three journal peer-reviewed paper as the first or second author and more than ten conference papers, and acted as a reviewer for top MR imaging journals.
Motion estimation and correction for fast MR imaging methods
MRI scanning is sensitive to subject motion due to lengthy data acquisition, which is longer than most types of physiological movement cycles. In order to reduce scanning time and impact from motion, parallel imaging methods have been introduced and applied in clinical applications. For example, the recently introduced wave-CAIPI achieves significantly lower g-factor penalty to image signal-to-noise ratio (SNR) than the traditional SENSE and GRAPPA methods. This method is capable to achieve a high acceleration factor and reduce the scan time of the whole-brain anatomical scan with a sub-millimeter resolution within 1 minute. Even though the scan time has been significantly shortened, significant motion-related image degradation using wave-CAIPI acquisition can persist, for example due to short sudden movement, in particular in the readout direction. This research is focusing on investigating the possibility of using joint optimization to estimate and correct motion for accelerated 3D wave-CAIPI acquisition.
Shawn Carere, B.Sc.Graduate Student, Medical Biophysics, University of Toronto
Shawn Carere is an MSc student in the Department of Medical Biophysics at the University of Toronto. He received his BASc in Electrical Engineering from Queen’s University with First Class Honours as well as was the recipient of the Ontario Professional Engineers Gold Medal upon graduation. Throughout his undergraduate degree Shawn gained relevant experience in machine learning and biomedical signal/image processing. He was an inaugural member, and later Director of Design, for QMIND, Canada’s largest undergraduate organization for artificial intelligence, now with over 250 members. He was also the the Technical Director for Merlin Neurotech, where he developed an API for real-time EEG signal processing. His capstone project, under the supervision of Dr. Shideh Kabiri Ameri, used machine learning to identify, classify and visualize heart arrhythmias in real-time using a portable ECG sensor.
Deep Image Reconstruction from Human Brain Activity using Capsule Networks
Shawn’s research is focused on designing new deep neural networks for reconstructing images from brain activity recorded with functional magnetic resonance imaging (fMRI). His current project intends to combine capsule networks, a novel alternative to CNN’s, and StyleGAN, a novel generator network, into a unique autoencoder network.
Angelica Manalac, B.Sc.Graduate Student, Medical Biophysics, University of Toronto
Angelica Manalac is an incoming MSc student in Medical Biophysics at the University of Toronto. She received her BSc in Medical Physics (Co-op) at McMaster University in 2020, with high distinction. During her undergraduate studies, she was a part of many different projects in the field of biophotonics. One project investigated the light propagation in breast tissue in order to obtain tissue optical properties and to characterize breast cancer risk. More recently, she was involved in executing Monte Carlo simulations of UVC light propagation for filtered face respirators to investigate their applicability for UV germicidal inactivation.
Angelica's MSc project will look at extracting characteristics of vasculature using gradient and spin-echo MRI.
Sayan Nag, B.Sc.Graduate Student, Medical Biophysics, University of Toronto
- Google scholar: https://scholar.google.com/citations?user=K8w4dj4AAAAJ&hl=en
- Personal website: https://sayannag.github.io/snag_site/
- Personal Art Gallery: https://sayannag.github.io/Gallery_Website/index-color.html
Effective Brain Connectivity for fMRI using Dynamic Causal Modelling
Experimental manipulations directly perturb neural activity, which is manifested in the fMRI response. In order to determine neuronal activity from experimental fMRI data, several biophysical generative graphical models like Dynamic Causal Model (DCM) and its variants have been proposed, which have great potential in computational psychiatry. These models can be employed to infer pathophysiological mechanisms from non‐invasively obtained measurements, which can guide differential prognosis and treatment prediction in individual patients. The physiologically-informed DCM (pDCM), which is developed in the lab of Dr. Uludag, is the state-of-the-art model. It is inspired by experimental observations about the physiological underpinnings of the fMRI signal to study effective connectivity in the brain. Along the same lines, I am using existing approaches such as pDCM and modifying them to study the effective connectivity in the brain on a bigger scale and dynamically.
Brian Nghiem, B.Sc.Graduate Student, Medical Biophysics, University of Toronto
Brian Nghiem is pursuing a Master’s degree in the Department of Medical Biophysics at the University of Toronto (U of T). He is interested in investigating applications of deep learning to carry out motion correction in MR neuroimaging. He received his Bachelor of Science degree at U of T in 2019, having completed the Biological Physics Specialist program with high distinction.
Outside of academics, Brian has maintained a life-long love for music. He is an active musician whose recent on-campus performances include the Women In Science and Engineering (WISE) 2019 Gala and occasional appearances with the Appassionata classical music group. He was previously a piano instructor with the Musical Minds Community Outreach program and the assistant director of the Innis College Choir at U of T.
Patient motion compromises the quality of MR images, resulting in artifacts that obscure potentially important features of the brain. State-of-the art retrospective correction methods attempt to estimate patient motion either by acquiring additional information during the scan time, or by solving a computationally-expensive optimization problem. More recent approaches involve the use of Deep Learning (DL) to improve motion correction by leveraging image priors learned during network training. Brian is focusing on the use of DL approaches to account for higher-order physics effects associated with motion. He is also interested in researching optimization methods and reference-free image quality metrics.
Jacob Schulman, B.Sc.Graduate Student, Medical Biophysics, University of Toronto
I’m a first-year master’s student in the department of medical biophysics at the University of Toronto, where I also received my undergraduate education in neurobiology and chemistry. Prior to my graduate work, my research experience (and interest) has spanned the intricacies of biological chemistry kinetics to DTI tractography/diffusivity analysis in paediatric Obsessive-Compulsive Disorder. I have had the privilege of being granted numerous academic scholarships for my undergraduate work. Outside of academia, I am a passionate singer/songwriter and guitarist, teacher and tutor, CrossFit enthusiast, and camp volunteer. I am very excited to work on clinical fMRI analysis with Dr. Uludag at BRAIN-TO.
Quantitative Perfusion Imaging with Dynamic Susceptibility Hypoxia Contrast
Visualizing blood flow in the brain is essential in the effective diagnosis, prognosis, and surgical mapping of neurological disorders, including cancer, neurovascular complication, and neurodegenerative disease. Dynamic susceptibility contrast (DSC) has proven to be an effective magnetic resonance imaging (MRI) technique for gaining qualitative and quantitative insight into blood flow in the brain (perfusion). A limitation to the DSC method is the use of an exogenous contrast agent, gadolinium, which has been known to accumulate in the brain and bones, and is associated with the development of nephrogenic systemic fibrosis in individuals with renal disease; thus, a further limitation to gadolinium contrast is an inability to image repeatedly within a relatively short time frame in such patients. Successive imaging is often critical for characterizing neurovascular impairment. A potential method to bypass gadolinium’s limitations comes in the form of a ‘gas challenge,’ a novel method which yields contrast by exploiting the inherent magnetic properties of oxygen’s carrier protein, hemoglobin. The gas challenge generates contrast by briefly and safely decreasing the oxygen content (temporary hypoxia), thereby increasing the concentration of deoxygenated hemoglobin (dHb). Since dHb is a paramagnetic substance capable of strengthening the local magnetic field in and around the blood vessels, the hypoxia model may produce a contrast for perfusion imaging.
The ultimate goal of my work is to determine the clinical efficacy of the hypoxia model as a safe alternative to gadolinium in DSC-MRI—to my knowledge, the first study of its kind. If validated, future work has the potential to benefit from and further explore this hypoxic model in applications of safer clinical imaging research and practice.
Labeeb Talukder, B.Sc.Graduate Student, Medical Biophysics, University of Toronto
Labeeb Talukder is currently pursuing a Master’s of Science degree in Medical Biophysics at the University of Toronto. He completed his honor’s bachelor’s of science at the University of Toronto Scarborough campus, receiving his Specialist Co-op degree in Neuroscience. During his undergraduate degree, a part of his research experience involved programming robotic exoskeletons for gait rehabilitation in patients with multiple sclerosis. This was done to assess the safety and efficacy for the clinical applications of technology originally licensed for military purposes. His undergraduate thesis involved studying audiovisual synchrony perception in infancy, utilizing infrared eye trackers to detect fixations patterns in response to synchronous and asynchronous audiovisual stimuli played at both high and low frequencies. He served as an emergency medical responder at the Scarborough Campus, a 24/7 campus-based first-aid response team. In his spare time, he enjoys staying active, with a particular interest in weightlifting.
His current research involves bridging fMRI research and artificial intelligence. Recent advances in computer vision have yielded the potential for neural networks to act as interpretive tools for understanding brain dynamics. This research will involve enhancing the biological plausibility of these neural networks and how we can use them as working models of the brain. Training these models on BOLD responses from fMRI has the potential to provide us with an unprecedented understanding of these underlying patterns of functional connectivity. A working model of the healthy brain will have important clinical applications for modeling neuropathological activity in the brain.
Project and Research Assistants
Paula Alcaide Leon, MDFull-time Clinician-Investigator, University of Toronto, and Staff Radiologist, JDMI, UHN, Toronto
Paula Alcaide Leon obtained her medical degree in 2005 at the University of Sevilla, Spain. Her diagnostic radiology training was also completed at the University of Sevilla from 2006 to 2010. She worked as a staff neuroradiologist in Barcelona in different institutions from 2010 to 2013 and completed a neuroradiology clinical and research fellowship at the University of Toronto from 2013 to 2016. From 2016 to 2019 she worked as a research specialist at the University of California San Francisco focusing on brain tumor imaging research. In 2020, Paula was appointed into a Full-time Clinician-Investigator position in the Department of Medical Imaging, University of Toronto and a Staff Radiologist position in the Joint Department of Medical Imaging at Toronto Western Hospital.
Paula has dedicated her academic efforts to ensure clinical translation of advanced MR techniques with the overriding aim of improving patient care and patient outcomes. Her research interests include brain tumor imaging, advanced imaging development and gender in healthcare.
Nadine Akbar, Ph.D.Postdoctoral Fellow (April 2020 - July 2020) LinkedIn Page
I completed my M.Sc. and Ph.D. through the Institute of Medical Science Department at the University of Toronto. The major focus of my graduate training was determining the neuroimaging (structural and functional MRI) correlates of cognitive function in persons with multiple sclerosis (MS). I then conducted two fellowships (Kessler Foundation in New Jersey, and then Queen’s University in Kingston) focused on the rehabilitation of fatigue in persons with MS. My fellowship at Kessler Foundation demonstrated how exercise training can induce changes to MRI functional connectivity of neural networks associated with cognitive fatigue. I have a continued interest in the use of neuroimaging tools to examine the neural mechanisms behind symptoms and rehabilitative interventions in various clinical populations.
My overall goal is to develop neuroimaging analysis tools that will help earlier detect and better treat cognitive, psychiatric and physical symptoms in various clinical disorders. We are at a time where there is a wealth of publicly available health-related and neuroimaging data that can be used by researchers to answer questions to better improve clinical care. My more specific goals are the development of analysis pipelines emphasizing the use of machine learning for single-subject structural and functional MRI prediction of cognitive and clinical variables in healthy and diseased populations using these large publicly available neuroimaging datasets.
Fatemeh RastegarMedical Biophysics Lab Rotation (April 2020 - May 2020)
Project: Evaluating a Python Reference Implementation of the conjugate-gradient SENSE Algorithm for MR Image Reconstruction
Lisa Rudolph, B.Sc.Research Technician (2019 - June 2020)
Lisa Rudolph is currently a research technician who received her B.Sc. in Honor’s Physics from the University of British Columbia in 2018. During her undergrad she cultivated a passion for particle physics, neuroscience, and computational physics – with the latter introducing her to machine learning. Her honor’s thesis delved further into the realm of machine learning as it consisted of creating and training an artificial neural network in order to optimize the cavity geometry of photonic crystals. Outside of work, she enjoys traveling, cooking, adventuring around Toronto.
Her latest project was focused on using machine learning to label fMRI images. Traditional methods for labelling an fMRI scan can be laborious, time-consuming, and prone to inconsistencies. Therefore, the alternative method she is looking at is using convolutional encoders-decoders to automatically label fMRI images by brain tissue type – a process known as semantic segmentation.