Radiomics and machine learning for the prediction of cardiovascular events
Lead Research Organisation:
University of Cambridge
Department Name: Medicine
Abstract
The impact of cardiovascular disease on society is enormous and yet, our ability to predict cardiovascular events at both the individual and population level remains poor. We will study the impact of radiomics and machine learning on risk prediction in cardiovascular disease. We hypothesize that using these two approaches will improve the classification of patients into high and low risk groups and improve prediction of clinical events, compared to Framingham and other contemporary approaches. Plan: the Cambridge datasets are labelled (e.g. with mode of clinical presentation) and will be divided into a training set and a validation set. Initially, image feature extraction will be used to identify regions of interest that can be combined together to build classifiers, ultimately to identify vulnerable atherosclerotic plaques (high-risk of rupturing vs. low-risk) and to predict outcomes. This approach can then be compared to a deep learning approach whereby a neural network model is developed and more complex, non-linear features in image classification can be learnt implicitly. The neural network develops from continual adjustments to its parameters and network architecture. Finally, with our collaborators in Cambridge we will apply radiomics and machine learning approaches to the vast UK Biobank dataset and the Addenbrooke's Hospital electronic health record (EPIC).
People |
ORCID iD |
James Rudd (Primary Supervisor) | |
Elizabeth Le (Student) |
Studentship Projects
Project Reference | Relationship | Related To | Start | End | Student Name |
---|---|---|---|---|---|
MR/N013433/1 | 01/10/2016 | 30/04/2026 | |||
1966157 | Studentship | MR/N013433/1 | 01/10/2017 | 30/09/2020 | Elizabeth Le |
Description | Supervised and Unsupervised Machine Learning in Cardiovascular Event Prediction |
Organisation | Imperial College London |
Department | Department of Computing |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | I attended reading groups and meetings to share ideas and discuss different methods to analyse the imaging and non-imaging data. I implemented the ideas that we discussed and experimented with different techniques to assess their performance on the dataset. |
Collaborator Contribution | We were kindly provided with access to a large dataset of well-characterised cardiac CT images, along with relevant clinical variables and follow-up data. Our collaborators were open to discussing ideas and different approaches to analysing the data and were helpful in providing support and teaching to acquire new skills in machine learning and programming. |
Impact | This is a multi-disciplinary collaboration involving clinicians and computer vision and machine learning experts. |
Start Year | 2019 |
Description | Supervised and Unsupervised Machine Learning in Cardiovascular Event Prediction |
Organisation | University of Edinburgh |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | I attended reading groups and meetings to share ideas and discuss different methods to analyse the imaging and non-imaging data. I implemented the ideas that we discussed and experimented with different techniques to assess their performance on the dataset. |
Collaborator Contribution | We were kindly provided with access to a large dataset of well-characterised cardiac CT images, along with relevant clinical variables and follow-up data. Our collaborators were open to discussing ideas and different approaches to analysing the data and were helpful in providing support and teaching to acquire new skills in machine learning and programming. |
Impact | This is a multi-disciplinary collaboration involving clinicians and computer vision and machine learning experts. |
Start Year | 2019 |