Periodic Reporting for period 2 - Neuro-Metrology (Brain connectivity metrology for personalised neuroimaging in health and disease)
Período documentado: 2023-04-01 hasta 2024-09-30
A key challenge towards untangling the complexity of the human brain is to understand how functionally specialised subunits are interconnected in the brain’s network to influence each other and produce experiences and behaviour. Magnetic resonance imaging (MRI) uniquely allows to explore this systems-level view of neural connections and to probe brain organisation. The connectome, the comprehensive map of brain connections, is unique in every person and reflects the organisation that constraints what the brain can and cannot compute. Yet, there are fundamental limitations in its mapping. Lack of standardised, accurate measures of brain connections through MRI and absence of objective references introduce errors and reduce interpretability and reproducibility.
Mapping and understanding brain connectivity is also fundamental for gaining insight into psychiatric disorders, which affect half a billion people worldwide (~165 million people in the EU alone) and according to the World Health Organization, they are amongst the leading causes of ill-health and years lived with disability. However, their diagnosis and treatment remain particularly challenging, due to a lack of understanding of how symptoms map onto brain circuits and potential abnormalities in the underlying neural connectivity. Therefore, having the ability to robustly measure disruptions of brain connectivity of individual patients can open up unique opportunities for aiding diagnosis and subsequently informing personalised treatment.
Building on my expertise, I will develop novel algorithmic platforms and image analytics for brain connectivity mapping. Contrary to conventional approaches that typically rely on ad-hoc processing, the NeuroMetrology team will establish measurement principles to allow, quantitative and objective characterisation of the brain connectome and its individual variability. Through a mixture of highly-interdisciplinary computational and experimental research, I propose a comprehensive framework governed by principles of metrology, the science of measurement. The project will develop methodologies that a) improve accuracy and precision of measurements of brain connections, by unifying diverse types of imaging information (WP1), b) harmonise and standardise such approaches for the individual across different measurement tools and MRI scanners (WP2) and link these measurements to reference standards, reflecting the population (WP3).
I will subsequently apply these new methodologies to investigate representative scientific questions that rely on the ability to capture personalised signatures of the brain architecture (WP4). Specifically, I will explore the links of neural connectivity patterns to behavioural traits and the potential associations of disruptions in brain circuits with symptoms in mental health disorders. In doing so, this programme will establish a paradigm shift in how neuroimaging data are analysed and neural wiring diagrams are obtained in a standardised manner, opening new translational possibilities for personalised medicine applications of neuroimaging-based phenotypes.
Mapping and understanding brain connectivity is also fundamental for gaining insight into psychiatric disorders, which affect half a billion people worldwide (~165 million people in the EU alone) and according to the World Health Organization, they are amongst the leading causes of ill-health and years lived with disability. However, their diagnosis and treatment remain particularly challenging, due to a lack of understanding of how symptoms map onto brain circuits and potential abnormalities in the underlying neural connectivity. Therefore, having the ability to robustly measure disruptions of brain connectivity of individual patients can open up unique opportunities for aiding diagnosis and subsequently informing personalised treatment.
Building on my expertise, I will develop novel algorithmic platforms and image analytics for brain connectivity mapping. Contrary to conventional approaches that typically rely on ad-hoc processing, the NeuroMetrology team will establish measurement principles to allow, quantitative and objective characterisation of the brain connectome and its individual variability. Through a mixture of highly-interdisciplinary computational and experimental research, I propose a comprehensive framework governed by principles of metrology, the science of measurement. The project will develop methodologies that a) improve accuracy and precision of measurements of brain connections, by unifying diverse types of imaging information (WP1), b) harmonise and standardise such approaches for the individual across different measurement tools and MRI scanners (WP2) and link these measurements to reference standards, reflecting the population (WP3).
I will subsequently apply these new methodologies to investigate representative scientific questions that rely on the ability to capture personalised signatures of the brain architecture (WP4). Specifically, I will explore the links of neural connectivity patterns to behavioural traits and the potential associations of disruptions in brain circuits with symptoms in mental health disorders. In doing so, this programme will establish a paradigm shift in how neuroimaging data are analysed and neural wiring diagrams are obtained in a standardised manner, opening new translational possibilities for personalised medicine applications of neuroimaging-based phenotypes.
We have developed novel approaches for mapping brain connections, showing for the first time how we can unify diverse data to jointly analyse and link very different brains. This allowed us to concurrently map for the first time areas that are uniquely human (phylogeny), as well as areas that develop later in humans (ontogeny) along with any common patterns in the phylogeny-ontogeny axis. We have also developed models where connections acted as scaffolds for propagation of neural activity, linking multiple modalities of neuroimaging (structural and functional) data. We have demonstrated how these models can be tuned to predict personalised patterns of brain activity. Subsequently, we have been developing novel approaches for estimating parameters for imaging-based models, and quantifying the parameter uncertainty using artificial neural networks.
We have built a comprehensive data resource for neuroimaging data harmonisation. By scanning the same participants across multiple imaging sites and scanners, we could establish how variable measurements of the same individual across MRI scanners are; how much of this variability is unwanted and caused by the measurement device and how much this interferes with true biological variability. We also used our data to investigate approaches for standardising the measurements, by ensuring that a good technique minimises the unwanted, scanner-induced variability and maximises information around biological variability. We considered hundreds of multi-modal imaging features and we identified which are more faithful and quantifiable across scanners.
Finally, we have built upon these developments to establish platforms for neuroimaging-derived personalised signatures. We have devised frameworks for robust brain-behaviour associations and for end-to-end integrative and scalable analytics. We will be using these platforms to analyse neuroimaging data of mental illness and explore how brain circuits and their disruptions link to symptoms.
We have built a comprehensive data resource for neuroimaging data harmonisation. By scanning the same participants across multiple imaging sites and scanners, we could establish how variable measurements of the same individual across MRI scanners are; how much of this variability is unwanted and caused by the measurement device and how much this interferes with true biological variability. We also used our data to investigate approaches for standardising the measurements, by ensuring that a good technique minimises the unwanted, scanner-induced variability and maximises information around biological variability. We considered hundreds of multi-modal imaging features and we identified which are more faithful and quantifiable across scanners.
Finally, we have built upon these developments to establish platforms for neuroimaging-derived personalised signatures. We have devised frameworks for robust brain-behaviour associations and for end-to-end integrative and scalable analytics. We will be using these platforms to analyse neuroimaging data of mental illness and explore how brain circuits and their disruptions link to symptoms.
For the first time, we have been able to study very diverse datasets and brains in a unified manner. We have demonstrated its utility in a challenging scenario of bringing together diverse data and mapping corresponding brain connections across species (humans and macaques) and across ages (neonate and adult human).
By bringing together a large team and multiple imaging sites, we have developed one of the most comprehensive travelling-heads multi-modal neuroimaging harmonisation resource to date. We have demonstrated how both implicit and explicit harmonisation approaches can be evaluated and developed using such a rich dataset, so that individual-specific imaging-derived phenotypes coming from different measurement devices (MRI scanners) are standardised, unbiased and precise. We have shown comprehensively for the first time the implications of denoising MRI data on bias and precision of derived measurements.
We develop data-driven artificial intelligence approaches a) for model inference and b) for normative modelling of neural connectivity features. We have working prototypes and we plan to further develop and investigate anomaly detection in pathology-induced disruptions.
Finally, we use the developed approaches in neuroimaging datasets from mental illness to explore how symptoms map to brain circuits and potential disruptions across individuals. We have already shown for the first time, how brain-behaviour and brain-symptoms association studies can provide unstable and non-generalisable multivariate patterns when poorly powered. We will build upon these findings in our further investigations that can subsequently open new ways for personalised diagnosis and treatment.
By bringing together a large team and multiple imaging sites, we have developed one of the most comprehensive travelling-heads multi-modal neuroimaging harmonisation resource to date. We have demonstrated how both implicit and explicit harmonisation approaches can be evaluated and developed using such a rich dataset, so that individual-specific imaging-derived phenotypes coming from different measurement devices (MRI scanners) are standardised, unbiased and precise. We have shown comprehensively for the first time the implications of denoising MRI data on bias and precision of derived measurements.
We develop data-driven artificial intelligence approaches a) for model inference and b) for normative modelling of neural connectivity features. We have working prototypes and we plan to further develop and investigate anomaly detection in pathology-induced disruptions.
Finally, we use the developed approaches in neuroimaging datasets from mental illness to explore how symptoms map to brain circuits and potential disruptions across individuals. We have already shown for the first time, how brain-behaviour and brain-symptoms association studies can provide unstable and non-generalisable multivariate patterns when poorly powered. We will build upon these findings in our further investigations that can subsequently open new ways for personalised diagnosis and treatment.