- Referencje
- University College Dublin
AI_PREMie aims to transform pre-eclampsia diagnosis with SAS® Viya® on Microsoft Azure
Research led by University College Dublin could save tens of thousands of lives using machine learning for new pre-eclampsia diagnosis
With 140 million births each year, more than 11 million women and their babies could be affected by pre-eclampsia globally. Difficult to diagnose and with complex origins, pre-eclampsia has a long-term impact on the health of mothers and their children. The condition claims the lives of 70,000 expectant mothers and 500,000 babies every year worldwide.
To tackle this public health challenge, Professor Patricia Maguire, Professor Fionnuala Ní Áinle and Professor Mary Higgins and their team at University College Dublin (UCD) launched AI_PREMie: an innovative research initiative that could transform pre-eclampsia diagnosis. It is designed to support clinicians to make effective and efficient clinical diagnosis of suspected pre-eclampsia through intelligent and timely risk stratification of patients’ clinical and blood biomarker data, using an advanced artificial intelligence (AI) algorithm. This helps improve health outcomes for women and infants.
Clinicians must have absolute trust in the system, and SAS and Microsoft were able to help us deliver on this right at the start. Patricia Maguire Professor University College Dublin
Supporting effective, efficient, clinical decision making
Every pregnant women with suspected pre-eclampsia will undergo multiple blood tests, but there is no effective rapid test to aid clinicians in diagnosis of pre-eclampsia. Research by the UCD team led to the discovery of important biomarkers with diagnostic and predictive power which, when combined with other clinical data, could be used to create a powerful AI (or machine learning) model to support clinical decision-making.
“We realised early on that we faced a scalability challenge with our solution,” comments Professor Maguire. “We could write our own code and develop our platform from scratch to combine and analyse all the data sets, but we would never be able to deploy it as we envisioned. We want AI_PREMie in every single hospital in the world and to make this test capability available to every single person who needs it. At that point, we turned to SAS and Microsoft to find out what would be possible.”
The team responsible for the data model is led by Ana Le Chevillier, Clinical Research Manager for Data and AI, closely supported by Dr Brian MacNamee, Associate Professor at the School of Computer Science, also at UCD. Ana comments: “We soon gained hands-on experience with the SAS Viya interface, building the data model with pre-eclamptic cases to train AI_PREMie. Our goal is to include at least 1,000 women in our training and validation test set from across Ireland.”
As more women are recruited to the AI_PREMie project, the team led by UCD will be able to refine the model. Currently, the solution data will be biased to the Irish population, and the team are aware that a global deployment would need a cohort to reflect the characteristics of the condition across demographic groups.
“Around five million babies are born prematurely every year due to complications of pre-eclampsia, with significant impacts on the baby and family, sometimes for the rest of their lives,” says Ana Le Chevillier. “The aim is to create an AI_PREMie consultant-in-your-pocket that provides expert clinical decision support at a critical time.”
University College Dublin – Facts & Figures
1 in 12
pregnancies affected by pre-eclampsia
5 million
premature births from pre-eclampsia complications
70,000+
maternal deaths and 500,000+ infant deaths caused by pre-eclampsia
Forging new partnerships
To take AI_PREMie to the next phase of development, UCD explored the option of building the solution using SAS Viya hosted on a Microsoft Azure cloud environment.
Professor Maguire remarks: “Both SAS and Microsoft were fully behind our mission and involved right from the beginning. They spoke with clinicians to work out the best way forward, and advise on how their technology could take the research from the lab to the clinical setting, allowing the team to focus on patient recruitment and model development.”
One major challenge was to design a solution that could be piloted successfully in a hospital once the research phase is complete, and then further scaled across multiple hospitals. “SAS and Microsoft brought experience from banking and government, for example bringing data security skills as well as scalability and resilience,” says Professor Maguire. “Clinicians must have absolute trust in the system, and SAS and Microsoft were able to help us deliver on this right at the start.”
Dr Katrina Comerford, Research Scientist at the School of Biomolecular Science, UCD, adds: “Our incredibly busy consultants made time to come to the weekly meetings with us, SAS and Microsoft. Diagnosing pre-eclampsia is challenging. The AI engine in SAS Viya enables us to analyse an increasing store of data points and biomarkers, and create a powerful decision support assistant.”
When clinicians are faced with complex individual cases, AI_PREMie will offer a pre-eclampsia risk score, with red, amber, or green flags to support their clinical decision; including a unified view of the patient data used to create the classification, incorporating unique platelet biomarkers as well as relevant clinical data and demographic information.
Dr Brian MacNamee, Associate Professor in UCD School of Computer Science, adds: “Having technologies like SAS and Microsoft will make it much easier for us to integrate with existing hospital systems. At the same time, the intuitive user interface of SAS Viya will help us to create a tool that clinicians can quickly feel confident using.”
To gain maximum benefit from the powerful SAS Viya software, UCD team members are enrolled in the free, online SAS STEP training program which helps biomedical and medical specialists develop new data science skills.
Ana Le Chevillier elaborates: “My original studies were in physiology and anatomy, and I have a non-coding background, yet SAS Viya is easy to understand and easy to use. We can show exactly which variables are going into the data, and it all makes sense to them (the clinicians). You can reach into the source data within SAS Viya in less than a minute, which provides the level of explainability that generates trust with the clinicians.”
My original studies were in physiology and anatomy, and I have a non-coding background, yet SAS Viya is easy to understand and easy to use. Ana Le Chevillier Clinical Research Manager for Data and AI University College Dublin
Helping to save lives
Currently, 500 pregnant women have been recruited to the project across three Dublin maternity hospitals: the National Maternity Hospital, the Rotunda Hospital and the Coombe Hospital. AI_PREMie is showing great promise. Looking ahead, UCD intends to directly connect with a patient’s Electronic Healthcare Record (EHR) to accelerate data ingestion.
Ana Le Chevillier comments: “Cutting edge capabilities from SAS and Microsoft Azure provide powerful interoperability and allow medical informatics exchange through Health Level Seven (HL7) and Fast Healthcare Interoperability Resources (FHIR®) standards, so it will offer us the ability to collect information direct from EHRs and integrate it directly into AI_PREMie. This will, in turn, enable us to scale rapidly and reliably in the future.”
With immediate plans to trial AI_PREMie in Dublin maternity hospitals later this year and roll out globally in the coming years, UCD hopes to help care providers around the world prevent thousands of deaths and medical complications caused by pre-eclampsia.
Professor Maguire concludes: “Every seven minutes an expectant mum loses her life because of the complications of pre-eclampsia. And about every 40 seconds, a little baby is lost again due to complications of pre-eclampsia. We’re looking forward to deepening our partnership with SAS as we develop AI_PREMie and, ultimately, help clinicians save more lives.”