2024-25 Project (Barrick & Howe & Isaacs)

Prediction of dementia from population-wide neuroimaging data using novel brain age techniques

SUPERVISORY TEAM

Supervisor

Dr Thomas Barrick at SGUL
Email: tbarrick@sgul.ac.uk

Co-Supervisor

Professor Franklyn Howe at SGUL
Email: howefa@sgul.ac.uk

Co-Supervisor

Dr Jeremy Isaacs at SGUL
Email: jisaacs@sgul.ac.uk

PROJECT SUMMARY

 Project Summary

This is an exciting opportunity to join a multidisciplinary translational magnetic resonance imaging research group that is developing and applying quantitative imaging biomarkers to determine brain age from neuroimaging data and use this to predict future incidence of dementia from UK Biobank data.    Alzheimer’s disease and vascular dementia are the most common dementias. Predicting individuals at risk of progression to dementia remains an unmet clinical challenge. Reliable well-validated biomarkers are required to predict long-term outcomes, monitor disease progression and objectively identify disease subtypes.    

The aim of this neuroimaging project is to use structural, tissue microstructural and functional biomarkers to develop multivariate machine learning and artificial intelligence algorithms to quantify the difference between brain age and chronological age to determine individuals at risk of dementia. This includes prediction of progression to mild cognitive impairment and dementia sub-types such as Alzheimer’s disease, vascular dementia and fronto-temporal dementia. Relationships of brain age with clinical and biological variables within the UK Biobank will also be investigated.    

This project will also apply novel tissue microstructural imaging techniques to UK Biobank data by application of Quasi-Diffusion Imaging (QDI), a novel, model-based, quantitative diffusion magnetic resonance imaging methodology developed at St George’s, University of London. QDI has not previously been applied to UK Biobank data and will enable quantification of new tissue microstructural biomarkers that are sensitive to ageing and specific to dementia.

Project Key Words

Magnetic Resonance Imaging, Quasi-Diffusion Imaging, Neuroimaging, Brain Age, Predicting dementia

MRC LID Themes

  • Global Health = No
  • Health Data Science = Yes
  • Infectious Disease = No
  • Translational and Implementation Research = No

Skills

MRC Core Skills

  • Quantitative skills = Yes
  • Interdisciplinary skills = Yes
  • Whole organism physiology = No

Skills we expect a student to develop/acquire whilst pursuing this project

  • Develop understanding of Alzheimer’s disease, vascular dementia and fronto-temporal dementia, and how neuroimaging may influence patient care and clinical decisions. 
  • Develop understanding of the physics of MRI, with emphasis on structural, diffusion and functional MRI techniques, and how neuroimaging may be used in general to influence patient care and clinical decisions. 
  • Develop novel imaging machine learning and artificial intelligence methods and apply statistical modelling methods to data. 
  • Gain experience in processing and analysing large multimodal datasets that include clinical, imaging, demographic and biological data. 
  • Develop understanding of the quasi-diffusion model of diffusion dynamics and quantitative tissue microstructural imaging biomarkers.  
  • Presentation of findings at clinical and academic conferences, in peer review publications and through public engagement.  
  • Understand challenges and opportunities of using patient data in health data science research.

Routes

Which route/s is this project available for?

  • 1+4 = No
  • +4 = Yes

Full-time/Part-time Study

Is this project available for full-time study? Yes
Is this project available for part-time study? No

Eligibility/Requirements

Particular prior educational requirements for a student undertaking this project

  • SGUL’s standard institutional eligibility criteria for doctoral study.
  • Minimum 2:1 honours degree.   
  • The ideal candidate will have studied at BSc or MSc level in one of: Computer Science, Physics, Engineering or Mathematics/Statistics or have a background in Magnetic Resonance Imaging or Neuroimaging.

Other useful information

  • Potential CASE conversion? = No

PROJECT IN MORE DETAIL

 Scientific description of this research project

1.       Background   
Alzheimer’s disease (AD) and vascular dementia (VD) are the most common dementias. Predicting individuals at risk of progression to dementia remains an unmet clinical challenge. Reliable well-validated biomarkers are required to predict long-term outcomes, monitor disease progression and objectively identify disease subtypes. Brain age methods quantify the apparent brain age of individuals from MRI data and the difference with chronological age, termed δ [e.g. 1,2,3,4]. Greater δ corresponds to accelerated brain aging and is a potential marker for brain degenerative diseases [1,2,5].   

Structural and tissue microstructural imaging techniques (using Diffusional Kurtosis Imaging , DKI) offer high sensitivity and specificity to identifying disease pathology in mild cognitive impairment (MCI) and dementia [6,7,8]. We have developed Quasi-Diffusion Imaging (QDI) [9,10,11] a technique that overcomes methodological limitations of DKI which will provide improved sensitivity and specificity to patterns of grey and white matter tissue microstructure in ageing and disease. QDI has not been previously used in δ quantification.     

2.        Project Objectives   
The project will use neuroimaging data from the UK Biobank. We hypothesise that:  (a) Increased δ will be associated with risk of MCI with greater δ corresponding to greater risk of dementia.  (b) δ enables prediction of progression to MCI and dementia.  (c) δ enables prediction of dementia subtype (AD, VD and fronto-temporal dementia (FTD)).     

3.          Techniques to be used and available data   
Several techniques will be used to quantify δ from neuroimaging data:  (a) Statistical techniques (i.e. Gaussian process regression [1,2,3,4]),  (b) Machine learning techniques (i.e.[1,4]),  (c) Artificial intelligence (AI) (i.e. neural networks [1,4]).  These techniques will be trained on quantitative neuroimaging features within anatomical grey and white matter (e.g. gyral, sulcal, deep grey structures and white matter bundles) including: (i) anatomical volumes (including vascular white matter damage), (ii) QDI tissue microstructural measures, (iii) functional MRI properties within cortical anatomy. Model development and training will be performed using 10-fold cross validation and include reduction careful modelling to reduce age-based biases [5]. Hypothesis (a) will investigate relationships between δ and clinical and biological variables.   

Machine learning and AI techniques will be developed to predict individuals progressing to MCI and dementia sub-types in Hypotheses (b) and (c). Sensitivity and specificity will identify the best techniques and features of structural, microstructural and functional anatomy associated with increased dementia risk will be identified.     

UK Biobank population-wide data (https://www.ukbiobank.ac.uk/, n=500000 participants, primary care data in n=230000) will be utilised within the project consisting of neuroimaging, clinical and biological data at cross-sectional baseline (n=100000) and 2 year follow-up (n=10000). Since inception of the UK Biobank project the individuals progressing to dementia have been identified (currently AD n=4451, VD n=2183, FTD n=319) indicating progression to AD (n=1935), VD (n=949), and FTD (n=139) for the cross-sectional imaging cohort and AD (n=194), VD (n=95), FTD (n=14) longitudinally.    4. Data availability statement    UK Biobank data (https://www.ukbiobank.ac.uk/) is available at a cost of £675 for the PhD project.   

5.          Potential risks to project and plans for mitigation   
To ensure development of brain age quantification software is achieved within the first two years of the studentship the QDI biomarkers will be computed within the first three months. To enable this Dr Barrick has developed QDI software for application to UK Biobank data.   

6. References   
[1] Franke & Gaser, Frontiers in Neurology, 10 (2019). 
[2] Cole et al., Molecular Psychiatry, 23 (2018). 
[3] Zhu et al., Translational Psychiatry (2023). 
[4] Dörfel et al., bioRxiv, (2023). 
[5] Smith et al., Neuroimage, 200 (2019). 
[6] Risacher et al., Current Alzheimer Research, 6 (2009). 
[7] Tu et al., Human Brain Mapping, (2021). 
[8] Raja et al., J Neurosci Methods, 335 (2020). 
[9] Barrick et al., Neuroimage, 211 (2020). 
[10] Barrick et al., Mathematics, 9(15) (2021). 
[11] Spilling et al., Magnetic Resonance in Medicine, 88(6) (2022).

Further reading

(Relevant preprints and/or open access articles)

Additional information from the supervisory team

  • The supervisory team has provided a recording for prospective applicants who are interested in their project. This recording should be watched before any discussions begin with the supervisory team.
    Barrick-Howe-Isaacs Recording
  • 10.1002/mrm.29420. Epub 2022 Aug 31

MRC LID LINKS