Multimodal longitudinal prediction of cerebrovascular disease
Lead Research Organisation:
UNIVERSITY COLLEGE LONDON
Department Name: Medical Physics and Biomedical Eng
Abstract
1) Brief description of the context of the research including potential impact
Cerebral small vessel disease (CsVD) is a chronic intracranial vascular disease that affects the small vessels and blood blow in the brain. The disease is prevalent in 28% of people aged over 80 and may lead to brain damage and cognitive impairment. The progression of CsVD is multi-factorial and multi-pathological. One important type of lesion in CsVD are white matter hyperintensities (WMHs).
Multiple etiologies have been proposed to explain the progression in WMHs. The Rotterdam Scan Study has found age, blood pressure, current smoking and the presence of lacunar infarcts as predictors of WMH progression after three years' of follow-up. APOE4 is another a potential marker. Interestingly, baseline WMH lesion load was found to be associated with WMH progression in the LADIS study, indicating the condition's accelerating progression rate. While WMHs are positively associated with age, the histopathology of WMH is heterogeneous. A prospective study found WMH regression in 37% of minor ischemic stroke patients after 1 year. As the load and locations of WMHs could determine their clinical relevance, those factors should be incorporated in future research. This could be enabled by an automated regional-zonal WMH burden quantification method that was previously proposed to describe the spatial distribution of WMHs over 4 layers and 9 lobar zones.
2) Aims and Objectives
The aim of the four-year project is to predict the evolution of cerebral small vessel damage based on multiple types of data and support early intervention. It is hypothesised that such multimodal approaches could provide a better indication of how WMHs may progress over time. The baseline work should include MRI only, and other sources of information should be added incrementally. These sources include other neuroimaging modalities, cardiac imaging,
and clinical data. The deliverables should provide uncertainty estimation alongside the regional predictions, and should evaluate the impact of additional sources of information on the prediction performance.
As a logical first step, the first-year project will focus on the prediction of WMH load at the regional level. In addition to the regional WMH volumetric measures, experiments could include an ablation study of different markers from other sources (eg. blood pressure). The study should be rigorously conducted, and exploit both internal validation and external validation using an independent dataset if time permits. As the regional progression is provided, effort should be taken to identify patterns of how the lesions shrink, grow or remain stable at different locations of the brain. In terms of the modelling approaches, references could be made to studies on other brain or heart conditions.
1. Data: We have identified the publicly available Alzheimer's Disease Neuroimaging Initiative 3 (ADNI3) cohort as the primary dataset. The two-centred data was collected between 2017 and 2023 (data collection still ongoing). In addition to the MRI scans, demographic information (eg. age, sex), medical history (eg. blood pressure) and biomarkers (eg. APOE4) are also available.
2. Model input: While the 4-year project is aimed at making longitudinal predictions based on MRI scans in 3D, the input to the MRes project is primarily regional WMH volumes from the ADNI3 cohort. As mentioned above, demographics such as age and sex, medical history such as blood pressure could be added as part of the experiments.
3. Expected outcomes: The anticipated outcomes are longitudinal predictions of WMH volumes at different locations of the brain.
3) Novelty of Research Methodology
The student will develop image-based multi-modal longitudinal prediction of white matter hyperintensities where research is currently lacking. Further work will extend to other types of CsVD.
Cerebral small vessel disease (CsVD) is a chronic intracranial vascular disease that affects the small vessels and blood blow in the brain. The disease is prevalent in 28% of people aged over 80 and may lead to brain damage and cognitive impairment. The progression of CsVD is multi-factorial and multi-pathological. One important type of lesion in CsVD are white matter hyperintensities (WMHs).
Multiple etiologies have been proposed to explain the progression in WMHs. The Rotterdam Scan Study has found age, blood pressure, current smoking and the presence of lacunar infarcts as predictors of WMH progression after three years' of follow-up. APOE4 is another a potential marker. Interestingly, baseline WMH lesion load was found to be associated with WMH progression in the LADIS study, indicating the condition's accelerating progression rate. While WMHs are positively associated with age, the histopathology of WMH is heterogeneous. A prospective study found WMH regression in 37% of minor ischemic stroke patients after 1 year. As the load and locations of WMHs could determine their clinical relevance, those factors should be incorporated in future research. This could be enabled by an automated regional-zonal WMH burden quantification method that was previously proposed to describe the spatial distribution of WMHs over 4 layers and 9 lobar zones.
2) Aims and Objectives
The aim of the four-year project is to predict the evolution of cerebral small vessel damage based on multiple types of data and support early intervention. It is hypothesised that such multimodal approaches could provide a better indication of how WMHs may progress over time. The baseline work should include MRI only, and other sources of information should be added incrementally. These sources include other neuroimaging modalities, cardiac imaging,
and clinical data. The deliverables should provide uncertainty estimation alongside the regional predictions, and should evaluate the impact of additional sources of information on the prediction performance.
As a logical first step, the first-year project will focus on the prediction of WMH load at the regional level. In addition to the regional WMH volumetric measures, experiments could include an ablation study of different markers from other sources (eg. blood pressure). The study should be rigorously conducted, and exploit both internal validation and external validation using an independent dataset if time permits. As the regional progression is provided, effort should be taken to identify patterns of how the lesions shrink, grow or remain stable at different locations of the brain. In terms of the modelling approaches, references could be made to studies on other brain or heart conditions.
1. Data: We have identified the publicly available Alzheimer's Disease Neuroimaging Initiative 3 (ADNI3) cohort as the primary dataset. The two-centred data was collected between 2017 and 2023 (data collection still ongoing). In addition to the MRI scans, demographic information (eg. age, sex), medical history (eg. blood pressure) and biomarkers (eg. APOE4) are also available.
2. Model input: While the 4-year project is aimed at making longitudinal predictions based on MRI scans in 3D, the input to the MRes project is primarily regional WMH volumes from the ADNI3 cohort. As mentioned above, demographics such as age and sex, medical history such as blood pressure could be added as part of the experiments.
3. Expected outcomes: The anticipated outcomes are longitudinal predictions of WMH volumes at different locations of the brain.
3) Novelty of Research Methodology
The student will develop image-based multi-modal longitudinal prediction of white matter hyperintensities where research is currently lacking. Further work will extend to other types of CsVD.
People |
ORCID iD |