06.05.2022

AI to predict cardiovascular disease based on fundus images

 AI to predict cardiovascular disease based on fundus images

Sechenov University’s focus on medical applications of artificial intelligence (AI) and machine learning is set to lead to a new tool that will facilitate the prediction of cardiovascular disease based on the images of the human eye fundus, with an accuracy of more than 90%.

The technology is expected to become available within 5–7 years. In the long term, fundus image analysis, used in the treatment of cardiovascular disease, could be a useful tool for cardiologists and ophthalmologists — because it will significantly reduce the time needed to assess disease progression and is much cheaper than conventional methods.

The fundus of the eye is almost the only area which can be used by doctors to analyse the small calibre vessels. The use of AI and machine learning for the analysis of abnormalities in fundus images can provide a most accurate prediction of the development of cardiovascular disease.

This research is being carried out at the World-Class Research Centre “Digital Biodesign and Personalised Healthcare”, a joint project of Sechenov University, Institute for Design-Technological Informatics of the Russian Academy of Sciences, Institute of Biomedical Chemistry (IBMC), Ivannikov Institute for System Programming of the Russian Academy of Sciences, and Yaroslav-the-Wise Novgorod State University.

The prototype software will be used at the Health Management Clinic of Sechenov University.

“In the long run, fundus image analysis — used for the treatment of cardiovascular disease — can become very helpful for ophthalmologists and cardiologists,” said Philipp Kopylov, Director of the Institute of Personalised Cardiology and Professor of the Department of Cardiology, Functional and Ultrasound Diagnostics (Sechenov University). “The main issue with accurate prediction of cardiovascular disease dynamics is associated — among other things — with the variety of data often obtained through complex and costly exams, such as coronary calcium scan, etc. Fundus image analysis is one of the ‘easy’ diagnostic methods. The software developed by us has already been able to ‘train’ the prediction algorithm with an accuracy of more than 90%. Now we are continuing to test it so we can reach a yet higher level of accuracy. This software — after completion of the research tests — will be subjected to clinical trials at Sechenov University.”

The World-Class Research Centre “Digital Biodesign and Personalised Healthcare” is a large joint initiative aimed at creating an effective system for predicting the development of oncology disorders and cardiovascular disease, based on the personalised, patient-oriented approach. The primary goals of this project are early diagnosis, remote patient monitoring, and creation of a digital biobank — a platform solution for collecting, storing, and processing functional, depersonalised patient data.