Developing network methodologies for disease classification

Theme (1)
The physiological network is an emerging concept in health data science. Its aim is to integrate a wide range of health-related data across the body to establish the quantification of a person's functional interdependencies, to provide important insights and nuanced classification strategies of different bodily states [1]. In this theme, I will pursue the integration of a physiological network approach with deep learning classification for large health datasets, focusing - at least in the first instance - on the UK Biobank. With this, I aim to develop critical novel methodologies for the advancement of understanding and classification of health and disease.

Theme (2)
It is of critical societal importance to develop methods for the detection of Alzheimer's Disease (AD) and other dementias in their early stages in the general population. AD is commonly known as a disconnection syndrome, whereby mass neuronal death leads to the disintegration of the brain's functional architecture [2]. I have developed novel complex network methodologies which have been developed and applied in pilot electroencephalography (EEG) data, finding abnormal topological and temporal functional patterns in patients with AD. I aim to build on this promising work, developing, refining and applying these methodologies to larger datasets of structural and functional imaging data (e.g., UK Biobank), building on my existing collaborations in the Alzheimer Scotland Dementia Research Centre. The information gathered will be used to analyse and classify AD in the general population.

References:
[1] Bashan, A., Bartsch, R.P., Kantelhardt, J.W., Havlin, S., Ivanov, P.C., Network Physiology reveals relations between network topology and physiological function. Nature Comms., 3: 702 (2011).
[2] Delbeuck, X., Van der Linden, M., Collette, F., Alzheimer's disease as a disconnection syndrome?, Neuropsychology Review, 13(2): 79-92 (2003).
[3] Bronstein, M.M., Bruna, J., LeCun, Y., Szlam, A., Vandergheynst, P., Geometric Deep Learning: going beyond euclidean data, IEEE Signal Processing Magazine, 34(4): 18-42 (2017).
[4] Smith, K., Escudero, J., The complex hierarchical topology of EEG functional connectivity, J. Neurosci. Methods, 276: 1-12 (2017).
[5] Smith, K., Ricaud, B., Shahid, N., Rhodes, S., Starr, J., Ibanez, A., Parra, M.A., Escudero, J., Vandergheynst, P., Locating temporal functional dynamics of visual short-term memory binding using graph modular Dirichlet energy, Scientific Reports, 7: 42013 (2017).

For objective (1) I will develop an integrated framework of complex networks and convolutional neural networks for health data which will build upon geometric deep learning [3], building established connections with Prof Pierre Vandergheynst of the École Polytechnique Fédérale de Lausanne (EPFL). Complex physiological networks will be derived from the multidimensional, complex datasets in the UK Biobank, which will then be utilised in this framework to gather important insights into health and disease and develop enhanced large-scale classification strategies.

For objective (2) I will build on pilot work (unpublished) which applies novel complex network methods [4,5] to detect AD from task-related EEG signals. This will be extended for the following:
(i) refinement and continued development of theoretical approaches to analyse functional connectivity
(ii) adaptations for application to high spatial resolution fMRI and MRI data
(iii) applied to resting-state fMRI in large datasets such as the Lothian Birth Cohort and UK Biobank
(iv) employment of powerful classification and regression approaches such as support vector machines

Such methods are also of great relevance to an array of different pathologies, including other forms of dementia, epilepsy, Parkinson's disease and schizophrenia, to which the strategy proposed will be generalised.

This project is in alignment with HDR UK priorities, developing novel data analyses and classification strategies for key national health problems, establishing national leadership, developing international collaborations, and strengthening interdisciplinary skills.

References: see Summary