Cardiac and vascular disease prevention after hypertensive pregnancy: insights from AI-derived, multi-organ, hypertensive disease progression models

Lead Research Organisation: University of Oxford
Department Name: RDM Cardiovascular Medicine


Women who develop blood pressure problems during pregnancy are more likely to have high blood pressure in later life as well as heart attacks or strokes. The children born to the pregnancy also tend to have higher blood pressure and are more likely to have problems during their own pregnancies. Our work has shown that the children and mothers have changes in their blood vessels, heart and brain that can be identified long before they develop high blood pressure or suffer the severe complications. We think these changes in the body develop slowly throughout their life and the progression of these changes is establishing their risk for later disease.

By understanding the pattern of changes across multiple parts of the body, over a lifetime, we think we can identify how advanced the underlying disease is for an individual and how their disease is likely to develop over the next few years. Furthermore, as the rate of change is likely to differ between parts of the body, and a change in one area of the body could drive development of disease elsewhere, we can also use this information to decide on optimal treatments. Certain treatments may be required to slow down the development of disease in a particular area and the selection of interventions may need to change at different stages of life and disease.

To test these ideas, we are going use large datasets we have acquired over the years based on imaging studies of women and children after a hypertensive pregnancy. The data includes information on the structure and function of several important organs such as the heart, brain and blood vessels. We will also expand these datasets by undertaking additional studies in older women, women who have grown up in different environments and in young adults. In all these studies, the same approaches have been used for the imaging and we will harmonise all the datasets so that we can study how this early underlying disease progression in different organs varies with their pregnancy history, age and their current health.

The initial analysis will focus on assessing changes in specific organs, but we also want to know how the patterns emerge across the whole body. To do this we will need to combine information from many different measures at the same time and use some of the latest advances in artificial intelligence (AI) to analyse the data collected in our studies, as well as other large studies of people within the UK. The computer will learn the multi-dimensional patterns of changes to the organs that occur as someone progresses from 'health' to a 'disease' state.

From this information we will discover the unique patterns of hypertensive disease development, and with that hope to open the door to better interventions and therapies tailored to each person. For example, specific stages of disease could be identified based on a particular combination of complex imaging markers. From this, we could then use the computer to learn the combination of simple measures that best approximate to the more complex pattern to generate tools that can be used in any hospital or community to manage and track disease development for an individual patient.

At the end of this programme of work we will understand how the body changes in mothers who have a hypertensive pregnancy, and their children, over the life course of their disease. This knowledge, enhanced by AI technology, has the potential to lead to simple tests that can be used in the clinic to determine the stage of disease for an individual. The results of these tests could then advise on the best preventive intervention (e.g. lifestyle choice) or treatment strategy that protect the woman or child from developing disease.

Technical Summary

Women who have a hypertensive pregnancy, and their children, are at significantly increased risk of cardiovascular disorders, including hypertension and stroke, in later life. Imaging studies show both the woman and child have identifiable cardiovascular changes decades before the development of these acute medical events.

We will study whether distinctive patterns in these imaging markers can be used to describe disease progression and, thereby, provide a clinical tool to identify optimal times and approaches to intervene and slow disease progression.

First, we will expand our existing post-partum imaging datasets to include older women and distinct demographic groups to form a harmonised dataset of complex, multi-modal, multi-organ phenotypic data extending from 6 months to 25 years post-partum. In addition, to understand early disease progression we will expand our existing young adult imaging datasets with additional follow up of a cohort at age 25-35 years with detailed pregnancy information.

We will then use novel unsupervised machine learning to learn the multi-dimensional disease progression landscape across heart, brain and vasculature related to hypertensive pregnancy in young adults and postpartum. Vascular disease patterns will then be compared to those more broadly related to hypertension identified within the UK Biobank imaging cohort using similar computational approaches.

Finally, we will investigate optimal ways to present this disease progression within clinical workflows to describe the relative position of individuals within the disease landscape. We will also determine whether models based on simple clinic measures can approximate to the complex imaging phenotypes.

At the end of this programme we hope to have used artificial intelligence to map the life course of disease progression in families affected by hypertensive pregnancy and identified simple clinical approaches to characterise and manage disease progression.


10 25 50