Anatomy-Driven Brain Connectivity Mapping

Lead Research Organisation: University College London
Department Name: Computer Science


The connectome, the comprehensive map of neural connections in the human brain, is unique in every individual. Even identical twins differ at the level of neural connectivity. Mapping the human connectome and its variability across individuals is essential in getting insight into the unknown cognitive aspects of brain function, but also into identifying dysfunctional features of the diseased brain.
For these reasons, understanding the human brain, its organisation and ultimately its function, is amongst the key scientific challenges of the 21st century. Magnetic resonance imaging (MRI) has revolutionised neuroscience by uniquely allowing both brain anatomy and function to be probed in living humans. Even if MRI allows only macroscopic features to be recovered (at the level of relatively large tissue regions, rather than individual neuronal cells), its non-invasive and in-vivo application has opened tremendous possibilities for brain research. Diffusion-weighted MRI (dMRI) is a particular modality that uniquely allows the mapping of fibre bundles, the underlying connection pathways that mediate information flow between different brain regions. The connection mapping is performed indirectly by processing dMRI images via computational algorithms referred to as tractography.
Tractography has already provided fundamental new insights into brain anatomy. The importance of brain connectivity to our understanding of the brain along, with the great potential revealed by tractography algorithms have led to large initiatives from both sides of the Atlantic. These utilise dMRI to collect state-of-the-art datasets of the healthy adult and the developing brain and map the structural connectome through tractography. They include the $30M NIH Human Connectome Project, the 15M Euros ERC Developing Human Connectome Project and the £30M UK funded Biobank Imaging. However, without state-of-the-art analysis methods, and new ways of analysing dMRI data, researchers will fail to get the most out of this vast wealth of upcoming data.
In this project, we propose new frameworks for tractography methods centred on neuroanatomy. We particularly focus on problems arising from ambiguous mapping of complex geometries (which are very common in the brain) to the dMRI measurements. These pose significant limits to the accuracy of existing approaches. We propose wholesale changes through computational and algorithmic solutions that will allow connections to be measured in-vivo with unprecedented detail, whole brain organization to be studied at a much finer scale and anatomical features -invisible to existing techniques- to be revealed. These advances will open new possibilities for neuroanatomical studies, but also set the foundations for new basic research in MRI processing and connectivity mapping. We will illustrate their potential using compelling demonstrator applications from basic and clinical neuroscience, including the assessment of benefits from using the new technology in assisting neurosurgical planning.
Description Two key findings so far:

- Using non-symmetric fibre orientation patterns improves brain connectivity mapping via tractography. Within the grant we have developed a method for estimating non-symmetric fibre configurations from MRI data using neighbourhood information and shown that embedding these new objects within existing tractography algorithms resolves long-standing ambiguities in brain connectivity mapping.

- Dispersion is often mistaken for crossing in diffusion MRI. Various studies suggest that fibre crossing configurations arise in a large proportion of MRI voxels in white matter of the brain. Within this grant we have challenges that view noting that more continuous patterns of fibre dispersion can appear as crossing although in fact should be considered separately. Experimental work reveals that many of the voxels previously identified as containing crossings in fact are more likely to contain dispersed fibre configurations. This influences the design of future tractography algorithms for brain connectivity mapping.
Exploitation Route These findings prompt further development of tractography algorithms and the underlying models they use, both of which are evolving fast.
Sectors Healthcare,Pharmaceuticals and Medical Biotechnology

Description Enabling clinical decisions from low-power MRI in developing nations through image quality transfer
Amount £1,020,000 (GBP)
Funding ID EP/R014019/1 
Organisation Engineering and Physical Sciences Research Council (EPSRC) 
Sector Public
Country United Kingdom
Start 02/2018 
End 01/2021
Description Image Quality Transfer
Amount £60,000 (GBP)
Organisation Microsoft Research 
Sector Private
Country Global
Start 10/2015 
End 09/2018
Description Learning MRI and histology image mappings for cancer diagnosis and prognosis
Amount £774,000 (GBP)
Funding ID EP/R006032/1 
Organisation Engineering and Physical Sciences Research Council (EPSRC) 
Sector Public
Country United Kingdom
Start 12/2017 
End 01/2020