A single dataset of up to 888,000 people—spanning diverse racial backgrounds—will be used to train artificial intelligence to spot subtle signs of heart failure in imaging and ECG scans before symptoms appear. Current genetic studies of heart structure have been too small and too narrow, mostly focused on white Europeans and basic measurements like heart size. This has left the genetic roots of heart failure largely unexplained. The project will extract both conventional metrics—chamber volume and pump function—and more sensitive markers, such as internal pressure changes, from heart images. By linking these measurements to genetic data, the team will search for new genetic regions that drive heart failure, then use Mendelian randomisation to test whether risk factors like obesity or high blood pressure actually *cause* the heart changes that lead to failure. If successful, the work could shift heart failure management from reactive treatment to personalised prevention. Doctors might one day calculate a person’s lifetime genetic risk score, then offer early lifestyle advice or pre-emptive medication—tailored to that individual and valid across ethnic groups. The findings could also point to drug targets for heart failure subtypes that currently have no effective treatments.
View original technical description
Heart failure is a common medical condition in which the heart does not efficiently pump blood around the body. It frequently leads to disabling symptoms and early death. Detection of abnormal changes in the heart structure and function by imaging aids heart failure diagnosis and can predict poor outcomes. Genetics determine variation in the heart structure and function as supported by the heritability assessment in family studies. Prior research into the genetic influence of heart structure and function focused on conventional heart measurements such as the heart size or overall pump function in a relatively modest number of mostly European (White) individuals. These research studies found a limited number of genetic loci (regions in the genome) explaining a small proportion of observed heritability in a population. This study will assemble one of the largest datasets containing heart imaging, electrocardiogram (ECG, a simple test to check heart rhythm), detailed lifestyle information and health outcomes in up to 888,000 individuals with diverse racial backgrounds. The large dataset will permit the development of reliable artificial intelligence (AI) techniques to automatically extract or estimate heart measurements from imaging and ECG. Conventional imaging metrics such as the volume of the heart chambers and the pump function as well as more advanced measures such as the pressure changes in the heart, which could be a more sensitive marker of a failing heart, will be developed. Using these imaging measurements along with individual genetic data, computational experiments will be carried out to discover the genetic markers which could provide additional knowledge on the origin and causes of HF and assist in the development of new medications which are urgently needed for certain types of heart failure. The results will also allow evaluation of the cause-and-effect relationships between cardiovascular risk factors such as obesity and high blood pressure and abnormal changes in the heart and eventual development of heart failure using a technique called "Mendelian Randomisation". This information will help in crafting targeted public health and preventive strategies. Lastly, with the data on genetic associations, we will build the genetic risk scores for each individual and test if they can predict heart failure development and perform well across different ethnic groups. If successful, this approach will transform the future management of heart failure by allowing doctors to provide a personalised assessment of each person's lifetime susceptibility to heart failure which will in turn permit early diagnosis, tailored lifestyle advice and pre-emptive medical treatment.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know