Periodic Reporting for period 2 - MENTALPRECISION (Integrating biology and behaviour for precision stratification of mental disorders)
Berichtszeitraum: 2022-12-01 bis 2024-05-31
In many areas of medicine, biological markers of disesase state have transformed diagnosis and treatment allocation. Psychiatry lags behind: disorders are still diagnosed by symptoms and no biomarkers have been found. However, addressing this is a formidable task because of a lack of analysis tools to understand the complex disruptions of mental disorders at multiple levels – from the level of brain circuits to behaviour. In addition, mental disorders are very heterogeneous at each of these levels. This means, for example, that different biological causes can result in the same symptoms and that two individuals with the same disorder can have very different symptom profiles.
The aim of this proposal is to provide a set of next generation analysis tools to solve this problem. More speficially, the project aims to stratify mental disorders on the basis of biological markers derived from brain imaging data derived from tens of thousands of people and quantitative measures of behaviour from smartphone-based monitoring.
We will build models to chart variation in multiple aspects of brain organisation across large brain imaging samples from more than 40,000 individuals based on ‘brain growth charting’ methodology we have developed. This will provide a platform to understand the shared and distinct mechanisms of mental disorders at the level of the individual. In addition, this project aims to develop innovative machine learning tools that will:
1. understand the dynamics of behavioural measures derived from smartphone monitoring
2. understand interplay between brain systems and the behaviours they underpin and how this gives rise to mental disorders
3. Integrate complementary information from behavioural and biological data
4. Stratify mental disorders in a way that cuts across diagnostic classifications and accommodates different mechanisms converging on the same symptoms.
These innovations will have far-reaching impact but in this project, we will apply them to predicting trajectories of resilience and risk in major depression and bipolar disorder which are a leading cause of worldwide disease burden.
The ultimate aim of this project is to develop a set of tools for psychiatry that will facilitate early, personalized intervention, preventative treatments and a better understanding of disorder entities.
The aim of this proposal is to provide a set of next generation analysis tools to solve this problem. More speficially, the project aims to stratify mental disorders on the basis of biological markers derived from brain imaging data derived from tens of thousands of people and quantitative measures of behaviour from smartphone-based monitoring.
We will build models to chart variation in multiple aspects of brain organisation across large brain imaging samples from more than 40,000 individuals based on ‘brain growth charting’ methodology we have developed. This will provide a platform to understand the shared and distinct mechanisms of mental disorders at the level of the individual. In addition, this project aims to develop innovative machine learning tools that will:
1. understand the dynamics of behavioural measures derived from smartphone monitoring
2. understand interplay between brain systems and the behaviours they underpin and how this gives rise to mental disorders
3. Integrate complementary information from behavioural and biological data
4. Stratify mental disorders in a way that cuts across diagnostic classifications and accommodates different mechanisms converging on the same symptoms.
These innovations will have far-reaching impact but in this project, we will apply them to predicting trajectories of resilience and risk in major depression and bipolar disorder which are a leading cause of worldwide disease burden.
The ultimate aim of this project is to develop a set of tools for psychiatry that will facilitate early, personalized intervention, preventative treatments and a better understanding of disorder entities.
To date, we have developed a set of tools for modelling brain imaging data at scale. This generalises the 'brain growth charting' techniques we developed prior to this project and enables them to scale to much larger datasets (from hundreds of subjects to tens of thousands of subjects) and to model more complex distributions. We have implemented these in software tools that we are have distributed to the community and we have also developed an easy to use software platform that enables scientists and clinicians to readily access these models without needing to write software of their own.
We have also brought models online that can be used to chart variation in multiple different aspects of brain structure and function. This includes the thickness and the area of the cerebral cortex, measures of brain volume and measures of the strength of functional connections between brain regions. These are all estimated using very large datasets (20-40,000 individuals). We plan to bring more models online as the project progresses
We have also developed analysis tools for smartphone monitoring data, although this work is still ongoing. This will enable us to understand how behaviours change over time, in the real world and provides the potential for detecting behavioural changes that could signal an imminent episode of psychiatric illness. We have written articles that put these techniques in the context of the emerging literature for smartphone-based monitoring of mental health.
We have also brought models online that can be used to chart variation in multiple different aspects of brain structure and function. This includes the thickness and the area of the cerebral cortex, measures of brain volume and measures of the strength of functional connections between brain regions. These are all estimated using very large datasets (20-40,000 individuals). We plan to bring more models online as the project progresses
We have also developed analysis tools for smartphone monitoring data, although this work is still ongoing. This will enable us to understand how behaviours change over time, in the real world and provides the potential for detecting behavioural changes that could signal an imminent episode of psychiatric illness. We have written articles that put these techniques in the context of the emerging literature for smartphone-based monitoring of mental health.
Our project advances beyond the state of the art in multiple ways:
1. The tools we have developed constitute a radical shift from conventional thinking in psychiatry, which has typically focussed on group averages. While this analytical approach is suitable for characterising how the 'average patient' from controls, it is not well suited to charting individual variation of the type necessary for clinical decision making. We provide principled methods to understand and map this variation, taking this further toward precision diagnostics by providing statistical tools to stratify cohorts whilst accommodating convergence of distinct mechanisms to on the same symptoms;
2. We provide unified models for individual differences integrating data from more than 50,000 individuals across the lifespan. This enables scientists and clinicians to bind disparate data to a common reference derived from population level data. This enables meaningful comparisons between different cohorts, populations and experimental paradigms. This will also advance clinical research by providing the ability to simulate clinical brain states, predict the effect of putative interventions and generate hypotheses using synthetic data.
3. We provide statistical and deep learning methods to model dynamics of measures of behaviour from smartphone monitoring. This can be used for many different smartphone platforms and to monitor behaviour in many clinical conditions, in real time and in the real world.
4. We will map the full range of variation –from health to disorder– across biological systems and behavioural dimensions whilst allowing deviations from purely dimensional features to capture distinct mechanisms and remaining agnostic to diagnostic labels and predefined behavioural dimensions and yet fully taking symptom- and illness trajectories into account.
5. We will contribute neuroimaging and smartophone monitoring data from the same individuals.
6. We provide multi-modal fusion techniques yielding a unified model for the emergence of symptoms from the dynamic interplay between environment, behaviour and brain systems.
This integrated, multi-level perspective sets this project apart from all current approaches. We will leverage the strengths of modern conceptualisations of mental disorders such as the U.S. Research Domain Criteria and the European Roadmap for Mental Health, whilst overcoming their limitations. We will also advance beyond recent reconceptualisations that consider psychopathology arising principally from interactions between symptoms, whilst neglecting biology. We consider that this project will have impact extending considerably beyond its lifespan
1. The tools we have developed constitute a radical shift from conventional thinking in psychiatry, which has typically focussed on group averages. While this analytical approach is suitable for characterising how the 'average patient' from controls, it is not well suited to charting individual variation of the type necessary for clinical decision making. We provide principled methods to understand and map this variation, taking this further toward precision diagnostics by providing statistical tools to stratify cohorts whilst accommodating convergence of distinct mechanisms to on the same symptoms;
2. We provide unified models for individual differences integrating data from more than 50,000 individuals across the lifespan. This enables scientists and clinicians to bind disparate data to a common reference derived from population level data. This enables meaningful comparisons between different cohorts, populations and experimental paradigms. This will also advance clinical research by providing the ability to simulate clinical brain states, predict the effect of putative interventions and generate hypotheses using synthetic data.
3. We provide statistical and deep learning methods to model dynamics of measures of behaviour from smartphone monitoring. This can be used for many different smartphone platforms and to monitor behaviour in many clinical conditions, in real time and in the real world.
4. We will map the full range of variation –from health to disorder– across biological systems and behavioural dimensions whilst allowing deviations from purely dimensional features to capture distinct mechanisms and remaining agnostic to diagnostic labels and predefined behavioural dimensions and yet fully taking symptom- and illness trajectories into account.
5. We will contribute neuroimaging and smartophone monitoring data from the same individuals.
6. We provide multi-modal fusion techniques yielding a unified model for the emergence of symptoms from the dynamic interplay between environment, behaviour and brain systems.
This integrated, multi-level perspective sets this project apart from all current approaches. We will leverage the strengths of modern conceptualisations of mental disorders such as the U.S. Research Domain Criteria and the European Roadmap for Mental Health, whilst overcoming their limitations. We will also advance beyond recent reconceptualisations that consider psychopathology arising principally from interactions between symptoms, whilst neglecting biology. We consider that this project will have impact extending considerably beyond its lifespan