Mapping brain network activity from structural connectivity using deep learning

Royal Commission Industrial Fellowship for Kellogg student Andrei-Claudiu Roibu with F. Hoffmann-La Roche Ltd

Talented young innovators from across the United Kingdom have been awarded prestigious Industrial Fellowships by the Royal Commission for the Exhibition of 1851, to develop solutions to some of society’s biggest challenges, from COVID-19 to climate change.

Set up by Prince Albert to organise the Great Exhibition of 1851 and extended in perpetuity to invest the profits in UK innovation, the Commission has been supporting promising research ever since. For 170 years it has provided crucial support to advance R&D and help to make UK industry more competitive internationally.

Through the Industrial Fellowships, the Commission brings together industry and academia to create commercially viable research and solutions for the mutual benefit of all. Fellows conduct their doctoral research with a company in their chosen industry, bringing academic expertise and approaches to a commercial operation. This enables students to investigate new ways of thinking about traditional problems, and forge exciting career opportunities. The programme also equips companies with cutting-edge research without the premium price tag and strengthens links between universities and commercial organisations.

Kellogg student, Andrei-Claudiu Roibu (DPhil in AI and Big Data Science applied to Neuroimaging) is investigating how deep learning and AI can be used to help model the relationship between the structure of the brain and its functional properties. In nature, the structure-function relationship is crucial, with an object’s shape enabling it to fulfil a job, as in the case of an animal’s body shape determining how fast it can run. Predicting the relationship between the structure and function of the brain is one of the primary goals of neuroscience. Achieving this will not only allow a deeper understanding of human neurobiology but also aid in the treatment of certain neurological conditions, such as Alzheimer’s or Parkinson’s. The raw data is obtained using magnetic resonance imaging (MRI) from the UK Biobank, a large long-term biomedical study and database containing information on tens of thousands of UK participants. MRI images and scans are invaluable for research, as they allow the study of the brain, its tissues, connections and activity in living subjects without the need for invasive operations or other medical procedures.

This AI model will learn the underlying relationship between the brain’s structure and its functions. Once working, the model can be expanded by adding other types of information, such as genetics and lifestyle data, not only to provide a better understanding of the structure-function relationship, but also to aid in practical applications such as neurosurgery and to gain a better understanding of rare neurological diseases, allowing for the development of new treatments. Not only does Andrei’s work have significant importance in the field of neuroscience, but also in the wider AI sphere, as the development of deep learning algorithms at this level of complexity could bring about contributions to both the fields of machine learning and medical imaging.