AI-based predictor of Type 2 Diabetes

Overview

Various news outlets reported over the Christmas 2024 period results of research around the use of Artifical Intelligence (AI) tools that can predict type 2 diabetes and pre-diabetes with a high degree of precision.

For reference, see:

BBC News – https://www.bbc.co.uk/news/articles/c80v1p5l4n1o

The Guardian – https://www.theguardian.com/society/2024/dec/23/nhs-to-begin-world-first-trial-of-ai-tool-to-identify-type-2-diabetes-risk

The research, funded by the British Heart Foundation (BHF), was recently presented at the American Heart Association’s (AHA) Scientific Sessions 2024, in Chicago. 

What is Type 2 Diabetes?

Type 2 diabetes is a chronic condition that occurs when the body doesn’t use insulin correctly, resulting in high blood sugar levels.

Symptoms – include increased thirst, frequent urination, tiredness, unexplained weight loss, and blurred vision. However, many people with type 2 diabetes show no symptoms, or symptoms are mild and take years to develop. 

Causes– type 2 diabetes is often linked to being overweight or inactive, or having a family history of the disease. 

Complications – over time, type 2 diabetes can lead to serious health problems, including damage to eyesight, feet, heart, and nerves. 

Treatment – type 2 diabetes may be treated via lifestyle changes like diet and exercise, medications, and regular check-ups. 

Early diagnosis is important to prevent the worst effects of type 2 diabetes. Around 5.6 million people in the UK are living with diabetes, and it is thought up to 1.2 million of those have the condition and are yet to be diagnosed.

Why is this a breakthrough?

Researchers at Imperial College London and Imperial College Healthcare NHS Trust believe that an AI tool that scans electrocardiogram results can identify the risk of type 2 diabetes as much as ten years before the patients begin to develop the condition.

See https://www.imperial.nhs.uk/about-us/news/ai-could-predict-type-2-diabetes-up-to-10-years-in-advance for reference.

An electrocardiogram (ECG) is a relatively quick and painless test that measures the electrical activity of the heart. It is the most common heart test and is used to check for heart problems.

According to the research team, 1.2 million ECGs from existing hospital records were used to train the tool named AI ECG Risk Estimation for Diabetes Mellitus (AIRE-DM).

AIRE-DM was then used to predict future risk of developing diabetes on samples of people of various demographic and socio-economic backgrounds. The results were correct in about 70% of cases.

AI therefore can unlock insights based on ECG results that the naked eye of human doctors may not always be able to spot. If the condition can be predicted well in advance of full development, patients at risk can be directed to prevention measures.

AIRE-DM’s predictive ability is further increased if ECG data is augmented with context about the patient: age, gender, blood pressure and weight history.

Next steps

According to Professor Bryan Williams of the British Heart Foundation:

We look forward to seeing how this technology could be incorporated into clinical practice, providing an opportunity to intervene early to help reduce risk and even prevent type 2 diabetes and its associated complications, altogether.

Two NHS hospital trusts in London – Imperial College and Westminster – have now started training AIRE-DM and clinical trials are planned for 2025.

As part of the trial up to 1,000 patients at both hospitals will have ECG scans read by the AIRE-DM.