I-AIM: Individualised Artificial Intelligence for Medicine
Lead Research Organisation:
University College London
Department Name: Computer Science
Abstract
Management and treatment of complex, chronic diseases such as Alzheimer's disease is one of the biggest challenges facing modern medicine. All clinical trials of investigational treatments for slowing or stopping the progression of Alzheimer's disease since 2003 have failed. This is likely due to the complexity and duration (decades) of Alzheimer's disease, coupled with the highly individual nature of the disease and its progression. Combined, this works against clinical trials by making it extremely difficult to identify and recruit a large group of individuals who are at the same stage of the same trajectory, and so who might benefit from a potential treatment. In principle, this challenge can be met by a set of modern computational approaches called data-driven disease progression modelling (D3PM), but some technological development is required first.
D3PM aims to combine statistics with the latest developments in AI and data science to estimate disease signatures that describe how a progressive disease plays out from beginning to end. This active research field grew from basic supervised machine learning (pattern learning/recognition) to a range of phenomenological (top-down) models, and mechanistic (bottom-up) models that incorporate a range of AI tools including unsupervised machine learning (pattern discovery). D3PM signatures have shown promise for estimating severity and predicting progression in neurodegenerative diseases such as Alzheimer's disease, but they currently lack in individual level precision, and mechanistic rigour.
This research and innovation project is a unique combination of technology development and translational product development: a series of novel technological developments for individualising D3PM and expanding mechanistic modelling; and translational efforts to develop drug-development tools based on this next-generation technology. In combination, this work will speed up drug-development by increasing the efficiency of clinical trials: recruiting smaller cohorts of suitable individuals will reduce costs and lead to fewer false-negative results - where a drug works on a fraction of the population, but the trial cannot detect it because the majority did not respond to treatment. The chosen application is Alzheimer's disease, but the ideas are fit-for-purpose for similarly complex, progressive diseases.
This fellowship is a significant launchpad for my career. My ambition is to benefit patients and society by providing robust computational solutions to complex healthcare challenges. My vision for achieving this ambition starts by targeting the global epidemic of dementia, where I have identified an unmet need (improving clinical trials) and proposed a viable solution in the form of this research and innovation project. The fellowship provides essential resources to capitalise on my recent progress in the field and to personally develop into a UK-based future leader in using AI for medicine and health.
D3PM aims to combine statistics with the latest developments in AI and data science to estimate disease signatures that describe how a progressive disease plays out from beginning to end. This active research field grew from basic supervised machine learning (pattern learning/recognition) to a range of phenomenological (top-down) models, and mechanistic (bottom-up) models that incorporate a range of AI tools including unsupervised machine learning (pattern discovery). D3PM signatures have shown promise for estimating severity and predicting progression in neurodegenerative diseases such as Alzheimer's disease, but they currently lack in individual level precision, and mechanistic rigour.
This research and innovation project is a unique combination of technology development and translational product development: a series of novel technological developments for individualising D3PM and expanding mechanistic modelling; and translational efforts to develop drug-development tools based on this next-generation technology. In combination, this work will speed up drug-development by increasing the efficiency of clinical trials: recruiting smaller cohorts of suitable individuals will reduce costs and lead to fewer false-negative results - where a drug works on a fraction of the population, but the trial cannot detect it because the majority did not respond to treatment. The chosen application is Alzheimer's disease, but the ideas are fit-for-purpose for similarly complex, progressive diseases.
This fellowship is a significant launchpad for my career. My ambition is to benefit patients and society by providing robust computational solutions to complex healthcare challenges. My vision for achieving this ambition starts by targeting the global epidemic of dementia, where I have identified an unmet need (improving clinical trials) and proposed a viable solution in the form of this research and innovation project. The fellowship provides essential resources to capitalise on my recent progress in the field and to personally develop into a UK-based future leader in using AI for medicine and health.
Planned Impact
I-AIM will impact the following important challenges related to Alzheimer's disease:
1. Discovery of precision treatments and novel treatment targets
Discovery of disease-modifying therapies in Alzheimer's disease in clinical trials has been hampered by heterogeneous cohorts of individuals at different stages of different trajectories within a disease, that obscures even strong effectiveness of a treatment in a subgroup. I-AIM will impact upon clinical trials by individualising data-driven disease progression modelling to provide clinical trials with homogeneous cohorts of individuals who are all at a similar stage of a similar trajectory, thereby empowering the discovery of precision treatments in the medium term.
I-AIM develops new computational models of disease mechanisms that reveal novel targets for future treatment development in the longer term.
Beneficiaries include patients (and future patients), pharmaceutical companies, clinicians, and society as a whole.
2. Precision patient care
I-AIM will enable precise characterisation of a patient's disease trajectory and their position on it. This precision empowers patient management and care decisions through, e.g., accurate prognosis, and differential diagnosis; which enables patients to be referred to get the support, and eventually the treatment, they need. These technological developments in I-AIM will impact upon the development of new workflows for clinicians informed by data-driven solutions.
Beneficiaries include patients and their families and carers, clinicians, and (long term) medical technology companies.
In the medium-to-long term, the technological developments in I-AIM promise considerable socio-economic impact through informing and improving clinical trials into putative disease-modifying therapies for Alzheimer's disease. Dementia, of which Alzheimer's disease is the primary cause, represents a global crisis in healthcare. Annual socioeconomic costs in the UK alone exceed £26 million (Alzheimer's Society 2014), including costs of care and lost productivity. Costs are rising rapidly with the ageing population. I-AIM's direct impact on drug-development efforts will decrease the mammoth socio-economic costs of dementia. For example, an Alzheimer's disease treatment that prolongs independent life by two years could save up to £12.9 billion per year by 2050
(Lewis et al., ARUK 2014; dementiastatistics.org). Such a treatment relieves untold personal suffering for vast numbers of Alzheimer's disease patients, their families, friends, and carers. There is also potential for vast economic impact to come from two revenue streams: increased productivity across society through prolonged independent life (and reduced burden on the healthcare system); and revenue from drugs licensed by the pharmaceutical industry. The latter revenue stream is expected to generate significant interest in I-AIM from the private sector, who will be engaged from the offset and throughout the project.
The scale of Alzheimer's disease and the novel, yet attainable, solutions proposed in I-AIM promise profound impact for society all the way from patients to the private sector and the wider public. These impacts are expected to entreat policy makers in Health and elsewhere to strongly consider data-driven AI solutions to complex problems.
I-AIM's impact on research and society will be expedited through wide communication and dissemination of results via open-access routes wherever possible and appropriate. This will be balanced by giving due consideration to commercialisation opportunities that will impact upon the private sector initially, such as drug-development tools, for which expert collaboration will be sought with the primary beneficiaries: pharmaceutical companies.
1. Discovery of precision treatments and novel treatment targets
Discovery of disease-modifying therapies in Alzheimer's disease in clinical trials has been hampered by heterogeneous cohorts of individuals at different stages of different trajectories within a disease, that obscures even strong effectiveness of a treatment in a subgroup. I-AIM will impact upon clinical trials by individualising data-driven disease progression modelling to provide clinical trials with homogeneous cohorts of individuals who are all at a similar stage of a similar trajectory, thereby empowering the discovery of precision treatments in the medium term.
I-AIM develops new computational models of disease mechanisms that reveal novel targets for future treatment development in the longer term.
Beneficiaries include patients (and future patients), pharmaceutical companies, clinicians, and society as a whole.
2. Precision patient care
I-AIM will enable precise characterisation of a patient's disease trajectory and their position on it. This precision empowers patient management and care decisions through, e.g., accurate prognosis, and differential diagnosis; which enables patients to be referred to get the support, and eventually the treatment, they need. These technological developments in I-AIM will impact upon the development of new workflows for clinicians informed by data-driven solutions.
Beneficiaries include patients and their families and carers, clinicians, and (long term) medical technology companies.
In the medium-to-long term, the technological developments in I-AIM promise considerable socio-economic impact through informing and improving clinical trials into putative disease-modifying therapies for Alzheimer's disease. Dementia, of which Alzheimer's disease is the primary cause, represents a global crisis in healthcare. Annual socioeconomic costs in the UK alone exceed £26 million (Alzheimer's Society 2014), including costs of care and lost productivity. Costs are rising rapidly with the ageing population. I-AIM's direct impact on drug-development efforts will decrease the mammoth socio-economic costs of dementia. For example, an Alzheimer's disease treatment that prolongs independent life by two years could save up to £12.9 billion per year by 2050
(Lewis et al., ARUK 2014; dementiastatistics.org). Such a treatment relieves untold personal suffering for vast numbers of Alzheimer's disease patients, their families, friends, and carers. There is also potential for vast economic impact to come from two revenue streams: increased productivity across society through prolonged independent life (and reduced burden on the healthcare system); and revenue from drugs licensed by the pharmaceutical industry. The latter revenue stream is expected to generate significant interest in I-AIM from the private sector, who will be engaged from the offset and throughout the project.
The scale of Alzheimer's disease and the novel, yet attainable, solutions proposed in I-AIM promise profound impact for society all the way from patients to the private sector and the wider public. These impacts are expected to entreat policy makers in Health and elsewhere to strongly consider data-driven AI solutions to complex problems.
I-AIM's impact on research and society will be expedited through wide communication and dissemination of results via open-access routes wherever possible and appropriate. This will be balanced by giving due consideration to commercialisation opportunities that will impact upon the private sector initially, such as drug-development tools, for which expert collaboration will be sought with the primary beneficiaries: pharmaceutical companies.
Publications
Aksman LM
(2021)
pySuStaIn: a Python implementation of the Subtype and Stage Inference algorithm.
in SoftwareX
Dekker I
(2021)
The sequence of structural, functional and cognitive changes in multiple sclerosis.
in NeuroImage. Clinical
Chen H
(2023)
Transferability of Alzheimer's disease progression subtypes to an independent population cohort.
in NeuroImage
Bron EE
(2022)
Ten years of image analysis and machine learning competitions in dementia.
in NeuroImage
Young AL
(2024)
Data-driven modelling of neurodegenerative disease progression: thinking outside the black box.
in Nature reviews. Neuroscience
Vogel JW
(2021)
Four distinct trajectories of tau deposition identified in Alzheimer's disease.
in Nature medicine
Leal GC
(2024)
Crop filling: A pipeline for repairing memory clinic MRI corrupted by partial brain coverage.
in MethodsX
Shand C
(2023)
Heterogeneity in Preclinical Alzheimer's Disease Trial Cohort Identified by Image-based Data-Driven Disease Progression Modelling.
in medRxiv : the preprint server for health sciences
Ravi D
(2022)
Degenerative adversarial neuroimage nets for brain scan simulations: Application in ageing and dementia.
in Medical image analysis
Davenport F
(2023)
Neurodegenerative disease of the brain: a survey of interdisciplinary approaches.
in Journal of the Royal Society, Interface
Oxtoby NP
(2022)
Targeted Screening for Alzheimer's Disease Clinical Trials Using Data-Driven Disease Progression Models.
in Frontiers in artificial intelligence
Young AL
(2021)
Ordinal SuStaIn: Subtype and Stage Inference for Clinical Scores, Visual Ratings, and Other Ordinal Data.
in Frontiers in artificial intelligence
Lopez SM
(2022)
Event-based modeling in temporal lobe epilepsy demonstrates progressive atrophy from cross-sectional data.
in Epilepsia
Scotton WJ
(2022)
A data-driven model of brain volume changes in progressive supranuclear palsy.
in Brain communications
Aksman LM
(2023)
A data-driven study of Alzheimer's disease related amyloid and tau pathology progression.
in Brain : a journal of neurology
Oxtoby NP
(2021)
Sequence of clinical and neurodegeneration events in Parkinson's disease progression.
in Brain : a journal of neurology
Marinescu Razvan V.
(2020)
The Alzheimer's Disease Prediction Of Longitudinal Evolution (TADPOLE) Challenge: Results after 1 Year Follow-up
in arXiv e-prints
Golriz Khatami S
(2022)
Comparison and aggregation of event sequences across ten cohorts to describe the consensus biomarker evolution in Alzheimer's disease.
in Alzheimer's research & therapy
O'Connor A
(2020)
Quantitative detection and staging of presymptomatic cognitive decline in familial Alzheimer's disease: a retrospective cohort analysis
in Alzheimer's Research & Therapy
Firth NC
(2020)
Sequences of cognitive decline in typical Alzheimer's disease and posterior cortical atrophy estimated using a novel event-based model of disease progression.
in Alzheimer's & dementia : the journal of the Alzheimer's Association
Bellio M
(2020)
Analyzing large Alzheimer's disease cognitive datasets: Considerations and challenges.
in Alzheimer's & dementia (Amsterdam, Netherlands)
Aksman L
(2020)
Tau-first subtype of Alzheimer's disease progression consistently identified through PET and CSF Neuroimaging: Understanding tau progression
in Alzheimer's & Dementia
Oxtoby N
(2020)
Piloting a novel screening tool for reducing heterogeneity in clinical trials in Alzheimer's disease Human/Trial design
in Alzheimer's & Dementia
Archetti D
(2020)
Inter-cohort staging efficacy of gaussian process progression model for Alzheimer's disease Neuroimaging / Optimal neuroimaging measures for tracking disease progression
in Alzheimer's & Dementia
Vogel J
(2020)
Spatiotemporal imaging phenotypes of tau pathology in Alzheimer's disease Neuroimaging: Understanding tau progression
in Alzheimer's & Dementia
Marinescu R
(2020)
Predicting Alzheimer's disease progression: Results from the TADPOLE Challenge Neuroimaging: Neuroimaging predictors of cognitive decline
in Alzheimer's & Dementia
Pascuzzo R
(2020)
Prion propagation estimated from brain diffusion MRI is subtype dependent in sporadic Creutzfeldt-Jakob disease.
in Acta neuropathologica
Description | Serious exploration of a startup business to exploit the technology. |
First Year Of Impact | 2021 |
Sector | Digital/Communication/Information Technologies (including Software),Pharmaceuticals and Medical Biotechnology |
Impact Types | Economic |
Description | Piloting A Secure, Scalable, Infrastructure for AI Dementia Research On Routinely Collected Data |
Amount | £157,985 (GBP) |
Organisation | Medical Research Council (MRC) |
Sector | Public |
Country | United Kingdom |
Start | 08/2022 |
End | 03/2023 |
Description | Prodromal dementia with Lewy bodies: Characterising the added value of structural MRI and computational modelling for differential diagnosis in a memory clinic (Pro-DLB MRI) |
Amount | £33,616 (GBP) |
Organisation | The Lewy Body Society |
Sector | Charity/Non Profit |
Country | United Kingdom |
Start |
Description | Transfer learning for clinical applications |
Amount | £20,000 (GBP) |
Organisation | Alan Turing Institute |
Sector | Academic/University |
Country | United Kingdom |
Start | 01/2021 |
End | 06/2021 |
Description | Bloomsbury Festival |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | Local |
Primary Audience | Public/other audiences |
Results and Impact | Hundreds of primary school students, plus general public over two days. A permanent display including active demonstrations and interactions with attendees at the Bloomsbury Festival in Bloomsbury, London. Lots of positive feedback from the schools and the festival organisers, concluding being invited back to the next year's festival. |
Year(s) Of Engagement Activity | 2021 |
URL | https://bloomsburyfestival.org.uk |
Description | DPM Tutorial at ISBI |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Professional Practitioners |
Results and Impact | A remote (due to CoViD-19) technical workshop on disease progression modelling at ISBI, the International Symposium on Biomedical Imaging. Attended by ~50 researchers from across the world. Lots of positive feedback and increased activity on our GitHub repository. |
Year(s) Of Engagement Activity | 2021 |
URL | https://disease-progression-modelling.github.io/pages/conferences/ISBI.html |
Description | DPM Tutorial at MICCAI |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Professional Practitioners |
Results and Impact | A remote (due to CoViD-19) technical workshop on disease progression modelling at MICCAI, the Medical Image Computing and Computer Assisted Intervention conference. Attended by ~50-100 researchers from across the world. Lots of positive feedback and increased activity on our GitHub repository. |
Year(s) Of Engagement Activity | 2021 |
URL | https://disease-progression-modelling.github.io/pages/conferences/MICCAI.html |
Description | SuStaIn Fest |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Professional Practitioners |
Results and Impact | A remotely run workshop on how to use our disease progression modelling software pySuStaIn. Has resulted in much activity on our GitHub repository and return interest from researchers looking to collaborate and/or run their own analyses using our algorithm. |
Year(s) Of Engagement Activity | 2021 |
URL | https://github.com/ucl-pond/pySuStaIn |