Associate ProfessorBen Fulcher

Project Opportunities

Title: Time-series biomarkers of neurological disorders

Summary of opportunity:

This research will develop a new machine learning framework for finding and quantifying patterns of brain dynamics that distinguish patients with brain disorders from healthy controls.

Opportunity synopsis:

Despite the ability of modern human brain-imaging technologies (such as EEG and fMRI) to produce incredible new data of our brain in action, scientists are yet to develop robust biomarkers for diagnosing and treating brain disorders. Methods for processing these data remain relatively crude, and analytical approaches have focused almost exclusively on how different brain areas communicate by quantifying pairwise statistical relationships between the activity dynamics of different brain areas (as ‘functional connectivity’). Our research suggests a way forward that would allow us to incorporate the activity dynamics of individual brain regions into our predictive models. This would allow us to understand where and what is different about brain dynamics in different disorders, with the potential to motivate new diagnosis and treatment protocols for debilitating brain disorders like schizophrenia.

--------------------------------------------------

Project Opportunities

Title: Highly comparative time-series analysis

Summary of opportunity:

This research involves developing new methods for time-series analysis based on a new analytic framework for understanding structure in time series.

Opportunity synopsis:

The world is constantly changing around us: from the fluctuations of the wind on our faces to the characteristic pulsing of our own heart-beat. How do we capture interesting patterns in these data that allow us to make useful predictions: What should we measure about heart beat dynamics to decide whether a subject is at risk of heart failure? What should we measure about economic data to predict whether the economy will grow or is at risk of collapse? What should we measure about a set of credit-card transactions to predict whether fraud is likely to be occurring? We have recently developed a unified machine-learning framework, hctsa, for leveraging thousands of scientific time-series analysis methods to partially automate solving problems of this type. The framework has seen wide success on a range of scientific problems, but much remains to be done! A broad range of projects exist and can be tailored to the specific interests of the student, including: (1) theoretical studies of criticality and other types of nonlinear dynamics, (2) solving specific real-world problems through data analysis via collaborations with clinicians and other scientists; and (3) properly calibrating our time-series feature library by developing simple empirical tests.

--------------------------------------------------

Project Opportunities

Title: Inferring the dimensionality of dynamical systems automatically using machine learning

Summary of opportunity:

This research will develop methods to infer the dimensionality of a dynamical system automatically, by adapting dimensionality reduction methods to high-dimensional time-series feature spaces.

Opportunity synopsis:

Finding simple principles that can explain the complexity of the world around us often relies on finding lower-dimensional representations (or manifolds) of high-dimensional data. The complex dynamics of many real-world systems can be well-approximated by the variation of only a handful of parameters. Successful inference of these parameters in a data-driven way would be a major advance with an impact felt across science and industry. For example, if we understand the hidden patterns underlying variability across patients with a given disease, doctors could learn what minimal set of tests to run on their patients in order to recommend the optimal treatment. In this project, the student will apply a new highly comparative analysis framework, hctsa, to time-series datasets to develop new data-driven methods to automatically infer the parametric dimensionality of a range of physical (and other real-world) systems.

--------------------------------------------------

Project Opportunities

Title: Modelling the mechanisms of brain stimulation

Summary of opportunity:

Brain-stimulation techniques that modulate brain activity in a targeted way are growing in their clinical relevance, for example, with transcranial magnetic stimulation (TMS) being an approved treatment for depression. However, despite its ubiquity in research and clinical contexts, we do not yet know how it works. We have recently shown how a mathematical model that describes interactions between neural populations in the cortex can reproduce the complex brain responses to stimulation, opening up an exciting new research avenue that uses quantitative analytical methods to guide groundbreaking new clinical applications.

Opportunity synopsis:

This project will develop a mathematical (neural field) model of how cortical circuits respond to brain stimulation. Methods developed in this project will be used to understand the physiology of brain stimulation, and to guide the development of individualised brain-stimulation protocols. Outcomes from this project will inform the development of the next generation of personalised TMS treatments. Students will collaborate with Dr Nigel Rogasch (University of Adelaide) to develop new mathematical models to explain their cutting edge TMS datasets. Travel opportunities are available.

--------------------------------------------------

Project Opportunities

Title: Multivariate brain activity patterns underlying consciousness

Summary of opportunity:

Clinical assessment of consciousness is one of the most significant issues in brain injury and general anaesthesia, yet it remains challenging for medical practitioners. Terrifyingly, up to 40% of brain-damaged patients who are assessed as unconscious are actually conscious. We desperately need an accurate consciousness measure based on clinically-measurable brain data, which would improve clinical care for patients under general anaesthesia and those who have suffered brain damage. In this project, we will develop new quantitative metrics for consciousness level from objective brain-imaging data.

Opportunity synopsis:

This work is in collaboration with an interdisciplinary Australian and international network of researchers working in consciousness theory and measuring state-of-the-art clinical datasets. Projects can be flexibly tailored to the students interests among the following tasks. (1) Develop a highly comparative multivariate time-series statistics that leverages the multivariate time-series analysis literature and the machine learning literature to find new objective measures of brain communication that are informative of consciousness level. This will include pairwise dependence measures (to deduce relationships between brain areas), and measures of distributed network structure (to quantify the structure of whole-brain communication); (2) Develop new quantitative measures of consciousness guided by the multivariate time-series analysis literature and inspired by a leading theory of consciousness: Integrated Information Theory (IIT); (3) assess different strategies against clinical neuroimaging datasets measured from collaborators at Melbourne University and around the world.

--------------------------------------------------

Project Opportunities

Title: Modeling brain dynamics with spatial gradients

Summary of opportunity:

A core aim of neuroscience is to understand how the brain, with its staggering complexity of 100 billion neurons, makes sense of the world around it. Large-scale global initiatives have measured the brain in intricate detail, but have given limited physical understanding of how the brain’s microscopic properties shape the whole-brain dynamics that underlie cognition. In this project, we will develop, constrain, and validate a new generation of physiologically based brain models that tightly integrate large-scale neuroscience data.

Opportunity synopsis:

Mathematical models of brain dynamics, based on physiology and physics, have been formulated and refined over decades, but they remain disconnected from the modern neuroscience data that have recently become available to the community. This project will bridge this gap, exploiting the wealth of intricate microscale brain maps and using it to refine our best neural field models of brain dynamics. This should allow us to connect the macroscale patterns observed in neuroimaging experiments (such as fMRI and EEG) to principles governing the interactions between large populations of neurons at the microscale. Progress would have dramatic consequences for understanding the healthy brain, and for diagnosing and treating brain disorders.

--------------------------------------------------

Project Opportunities

Title: Algorithmic principles of navigating the brain using nanorobots

Summary of opportunity:

Over the past several years, high-throughput neuroscience methods have yielded comprehensive cellular maps of the entire brain. These data have revealed features, like distinctive marker cells and low-dimensional molecular patterns, that could be exploited by nanoscale devices to enable efficient navigation. At the same time, nanoscale sensing devices, such as those fabricated using DNA origami, are becoming increasingly sophisticated in their ability to perform useful computations. Accurate nanoscale navigation would enable transformative new treatments of brain diseases and cognitive therapies via targeted drug delivery.

Opportunity synopsis:

In this project, we will perform new physics-based characterizations of molecular patterning in the brain, and use physical simulations to investigate how nanoscale sensing rules can efficiently navigate to a given target location in the brain. The student will work with whole-brain neuroscience datasets, and use methods from statistical learning and physics to develop optimal sensing rules required for accurate and efficient brain navigation. This project is supported by a diverse team as part of the School of Physics Grand Challenge in Nanoscale Brain Navigation.

--------------------------------------------------

Project Opportunities

Title: Analysing non-stationary complex systems with applications to sleep

Summary of opportunity:

This scholarship is to support an Australia Research Council funded project on data-driven methods for tracking non-stationarity in time series, with applications to sleep data. The student should have strong background experience in quantitative methods (preferably with training from physics or mathematics and experience with data-driven analysis), and a keen interest in time-series analysis and sleep.

_x000D_

 

Opportunity synopsis:

--------------------------------------------------

Project Opportunities

Title: Analysing non-stationary complex systems with applications to sleep

Summary of opportunity:

This scholarship is to support an Australia Research Council funded project on data-driven methods for tracking non-stationarity in time series, with applications to sleep data. The student should have strong background experience in quantitative methods (preferably with training from physics or mathematics and experience with data-driven analysis), and a keen interest in time-series analysis and sleep.

Opportunity synopsis:

----------Current Research Projects----------

_x000D_ Our group is working across two main areas:

_x000D_

_x000D_ Dynamics. We are working to develop new general analysis methods to understand the dynamics of complex systems, including feature-based time-series analysis, multivariate information dynamics, machine learning, and dimensionality reduction.

_x000D_

_x000D_ Neural Systems. We are working to develop new ways to understand the structure and function of neural systems using physical and statistical modelling. We work with experimentalists to develop and refine mathematical models of whole-brain activity dynamics, and use statistical methods to find new biomarkers for brain disorders.