Developing synergies between transient astronomy and early medical intervention

Lead Research Organisation: University of Portsmouth
Department Name: Institute of Cosmology and Gravitation

Abstract

The application of artificial intelligence (AI) to healthcare will revolutionise medicine and patient care over the coming decades. The wealth of data being collected by medical professionals keeps growing and they need significant help to quickly understand this information for their key diagnostic tests. In particular, can this wealth of digital data be used to automatically detect possible problems early, and thus alert doctors for closer inspection?

Astronomy is also experiencing a data revolution driven by STFC experiments like LIGO and LSST. These experiments are opening up the time domain in astronomy and in the near-future astronomers will be deluged by millions of transient events a day; LSST will produce on average hundreds of transients per second! The prioritisation of so many events will be key to our success and astrophysicists are already developing and applying AI technology to help find the most interesting events within the ocean of "ordinary" transients.

This project brings these two problems together by applying the knowledge and expertise developed in transient astronomy to the early automated diagnosis of medical problems. This project is at the forefront of interdisciplinary research bring world-leading cardiologists from King's College Hospital together with the best transient astrophysicists in the UK. Together, we hope to find early diagnostic signatures in detailed time-series patient data e.g. ECGs. This pilot activity will be used to prime-pump future investigations and grow the number of scientists engaged in such impactful research.

Planned Impact

This proposal will provide a range of impacts:

1. Development of interdisciplinary research on two cardiology data science projects to create early AI diagnostic tests for cardiac arrest. While still at an early stage, we would hope the outcomes of these projects would inform and influence long-term healthcare providing quicker alerts for doctors to make time critical decisions. Such societal impact is at the heart of the UK Industrial Strategy which seeks to put the UK at the forefront of AI in healthcare.

2. We will engage with a larger range of scientists, doctors and industrialists across the southeast to grow research and innovation capacity for precision healthcare and early diagnosis. These innovators will then be encouraged to explore their ideas through two dedicated hackathons provide rapid prototyping of possible solutions. These solutions will form the basis for future proposals and investment, gain both societal and commercial impact.

Publications

10 25 50
 
Description Our use of astronomy algorithms and techniques can help advance understanding and interpretation of medical data through collaboration with doctors and clinical staff. Their expertise is vital in the design of any advanced techniques while astronomers provide the know-how to make it work.
Exploitation Route We are preparing publications in medical jorunals on our collaboration.
Sectors Healthcare

URL https://www.ukri.org/news/star-spotting-science-used-to-spot-cancerous-cells/
 
Description https://www.ukri.org/news/star-spotting-science-used-to-spot-cancerous-cells/
First Year Of Impact 2020
Sector Healthcare
Impact Types Societal

 
Title KOCAR Dataset 
Description We have a high quality and cleaned dataset of around 400 patients who have suffered Out of Hospital Cardiac Arrest. This database has formed the foundation of the research we have conducted as part of this award. 
Type Of Material Database/Collection of data 
Year Produced 2018 
Provided To Others? No  
Impact This dataset has resulted in the KOCAR web tool (described in another researchfish section), and also a publication which has been submitted for review. 
 
Description King's College Hospital Collaboration 
Organisation King's College Hospital
Country United Kingdom 
Sector Hospitals 
PI Contribution We have used statistical analysis methods commonly used in astronomy to analysis the patient data provided by King's College Hospital. We have worked closely with the Department of Cardiology to advise on data analysis and collection techniques to facilitate future collaboration.
Collaborator Contribution Our collaborators have provided data to our group for research and also co-supervise placement students.
Impact We have submitted a paper to a cardiology journal that explores a decision-tree based method for the classification of a culprit coronary lesion in cardiac arrest patients.
Start Year 2017
 
Description Kings College Hospital Trust 
Organisation King's College Hospital NHS Foundation Trust (NCH)
Department Cardiology
Country United Kingdom 
Sector Academic/University 
PI Contribution Expertise and skills in data science based on astronomical data analysis techniques
Collaborator Contribution Access to unique cardiology data including clinical expertise in interpreting any data analysis results
Impact results still being analysed but looking profitable to find better treatments especially for emergency medicine
Start Year 2018
 
Description Molegazer 
Organisation Oxford University Hospitals NHS Foundation Trust
Country United Kingdom 
Sector Academic/University 
PI Contribution Arising from a Cancer ResearchUK sandpit event the Molegazer collaboration aims to develop early detection of melanoma from time-series analyses of naevi evolution using techniques developed in super nova astronomy
Collaborator Contribution The School of Astronomy and Physics at Southampton are collaborators with us on Super Novae research. Oxford University Hospital Trust are the problem owners and have provided images for initial analysis as well as clinical knowledge.
Impact Proposals to Welcome Trust and CRUK both unsuccessful
Start Year 2019
 
Description Molegazer 
Organisation University of Southampton
Country United Kingdom 
Sector Academic/University 
PI Contribution Arising from a Cancer ResearchUK sandpit event the Molegazer collaboration aims to develop early detection of melanoma from time-series analyses of naevi evolution using techniques developed in super nova astronomy
Collaborator Contribution The School of Astronomy and Physics at Southampton are collaborators with us on Super Novae research. Oxford University Hospital Trust are the problem owners and have provided images for initial analysis as well as clinical knowledge.
Impact Proposals to Welcome Trust and CRUK both unsuccessful
Start Year 2019
 
Title KOCAR Web App 
Description The KOCAR web tool provides an easy method of inputting data into the decision tree-based model we developed as part of this award. The web app asks the user a series of Yes/No questions to determine the probability of a patient having a culprit coronary lesion. 
Type Of Technology Webtool/Application 
Year Produced 2019 
Open Source License? Yes  
Impact The web app is currently used retrospectively until the paper is published and the tool approved. It has the potential to guide patient treatment. 
URL https://kocar-250409.appspot.com/