Coronary heart disease and stroke are the world’s biggest killers, responsible for 16% and 11%, respectively, of global deaths, according to the WHO. Fighting against these conditions is the aim of the Dallas-based American Heart Association (AHA), one of the largest non-profit organisations in the world spearheading cardiovascular medical research.
At the core of this research is AHA’s Institute for Precision Cardiovascular Medicine’s Precision Medicine Platform – a cloud-based environment that allows the global medical research community to share huge amounts of datasets effectively and is used to develop algorithms that help predict cardiovascular disease earlier. It also enables researchers to access AI and machine learning-powered tools, namely TensorFlow, Keras, NiLearn, PyTorch, RandomForest, C50 and Caret, that can quickly analyse data and inform prognoses and treatment.
Dr Jennifer Hall, chief of data science and director of the Institute for Precision Cardiovascular Medicine at AHA, spoke to Tech Monitor about how the platform is making precision cardiovascular research more efficient and collaborative.
Platform to democratise data
As one of the biggest funders of scientific research in the US, AHA has clinical registries with over 10 million records from more than 2,500 hospitals across the country. Hospitals are encouraged to join this clinical registry programme – called ‘Get with the guidelines’ – so they can access the data to understand how they can improve the quality of care of patients in different areas, including atrial fibrillation, stroke, heart failure, coronary artery disease and now Covid-19 too.
“This data helps us understand how to best treat patients,” says Hall. “What’s the best treatment for a patient with hypertrophic cardiomyopathy? What’s the best treatment for patients with stroke?”
Before the platform was built in 2016, data was siloed in local labs, and many hospitals did not have the latest neural network and machine learning tools, or the specialist staff, required to analyse it. The Precision Medicine Platform bypasses these obstacles by making the data available to the wider healthcare community and relevant industry stakeholders. It also equips researchers and those accessing the platform with the necessary tools to access and manage the data.
Built on an AWS back-end, it was founded on what is dubbed the FAIR data principles: data that is findable, accessible, interoperable and reusable. “We are making it [data] available to everyone, not just people that have resources, not just individuals that are at the top universities or those that know how to code and all that, but to all businesses and people that can understand [that data] from different points of view,” explains Hall.
All the records that are part of the clinical registries are subject to the strictest security and privacy requirements. “Security is our number one concern that we account for in many ways,” says Hall. “We call this data governance – and it puts the control for data sharing in the hands of the custodian of the data.” When someone requests data, they need to fill out a form that is then emailed to the data custodian, who will then review this request before coming up with a decision. If the form is approved, the data is then moved into the private workspace of the data requester.
Despite these security protocols, Hall says that there is an ongoing resistance among people in academia and industry that still have the mindset that cloud security is not as good as on-premise – something that will take some time to change.
Precision medicine: predicting heart disease
The cloud-based platform also provides private workspaces where users can access and analyse data using AI-enabled tools. For example, by applying machine learning to angiograms, it is able to find out which patients need an urgent coronary stent. It can also repurpose three million MRI images to improve artery blockage and cardiovascular risk predictions. The process time for analysing these images has been reduced from four hours to two minutes thanks to the platform, AHA says, and it costs a fraction of what traditional analysis techniques did.
The data collected in the platform also functions as a knowledge library that researchers can use to understand risk models for heart failure and stroke – findings that are then published and disseminated among the wider scientific community. The platform also incorporates other information, such as social and educational data, that can help doctors understand patients’ risk.
“Today, our team makes it very easy to utilise these tools – state of the art sort of statistical analysis tools that you plug and play,” says Hall. “Essentially, we put the tools in the hands of the people that need it, and make it easy for the end user to look at all the data and make new conclusions about the data to improve patient lives.”
During the pandemic, Hall says that AHA also established a Covid-19 registry in less than a month, becoming one of the largest in the US with over 35,000 individual records. Through the Precision Medicine Platform, researchers and other stakeholders can access published data and disseminate new findings that have changed the way healthcare professionals envision the risk of Covid-19 and cardiovascular disease.
Last week, AHA’s Precision Medicine Platform entered an alliance with data management and analytics provider Hitachi Vantara to accelerate the platform and improve its accessibility for healthcare organisations. The company has funded over 90 data scientists and engineers to improve and maximise the platform’s tools.
“That’s what makes it so unique and special: to meet the needs of the healthcare sector, because these data scientists and engineers have been in that workspace,” Hall concludes. “They know what their clinicians need; they know what is needed.”
