Artificial Intelligence to improve Cardiometabolic Risk Evaluation using CT (ACRE-CT)

Research summary

The diabetes global epidemic affects 8% of the world's population and causes spiralling healthcare costs. Fat tissue (called adipose tissue) can become inflamed. This occurs differently in various parts of the body. This inflamed tissue can be a critical factor in the major complications of diabetes,such as cardiovascular disease (CVD). But not all fat tissue is considered 'harmful'. The risk factor relates to where in the body it is situated,and the biological characteristics of the fat. Even in people who are not obese,inflamed fat ‘hidden’ in certain locations can still be a risk factor. Currently,none of the routine scans or tests can identify inflammation in fat tissue. This study will focus on how best to detect this tissue using routine scans. Our new method,called FatHealth,detects fat tissue inflammation using new artificial intelligence techniques applied to routine computed tomography (‘CT’) scans. FatHealth can identify people who may be at risk of developing diabetes,and also people with diabetes who are at high risk of death from cardiovascular disease. This new method is better than other diagnostic tests for this purpose. Caristo Diagnostics in association with the University of Oxford will collaborate with other institutions to transform FatHealth into a commercial product. The intention is to analyse 20,000 CT scans to train its AI algorithm. We plan to develop an automated web-based system for rapid and reliable AI analysis of CT scans to deliver accurate risk predictions. We will work with clinical NHS organisations and patient groups to evaluate the effectiveness of the FatHealth test in patients at risk of diabetes. We will conduct a clinical study that compares the results of FatHealth to the current method for diagnosing diabetes and pre-diabetes (the Oral Glucose Tolerance Test).

Principal Investigator

Prof Charalambos Antoniades

Contact us

Email: cvm_nurses@cardiov.ox.ac.uk

IRAS number

303226