Algorithms of computer vision for coronary health

The group BCN Perceptual Computing Lab at the Faculty of Mathematics of the UB has developed a computer application that studies IVUS images and facilitates their interpretation and quantifying. Researchers, led by the UB professor Petia Ivanova Radeva, have worked together with a group from the Hospital Germans Trias i Pujol, managed by Doctor Josepa Mauri, and the North-American company Boston Scientific, worldwide leader in coronary interventional products. The firm has already set up the process to include the computer application created by the UB group and the Hospital in its products. “To know that some of our algorithms will provide worldwide doctors with more interventional tools to improve patientsʼ health is absolutely motivating for us”, explains Petia Ivanova Radeva, from the Department of Applied Mathematics and Analysis of UB.

The Board of Trustees and the Bosch i Gimpera Foundation conferred the Antoni Caparrós Prize for the best project of knowledge transfer on the collaboration among UB, the Hospital Germans Trias i Pujol and Boston Scientific, which started in 2005 and was divided in two stages. In the first one, researchers from the University designed and developed some algorithms based on computer vision and machine learning to detect and quantify the contours of coronary arteries. It is necessary to analyse blood vessel contours in order to observe the presence of plaques formed by cholesterol, calcium and fibrous tissue, which may cause thrombosis or coronary ischemia. In the second stage of the project, an innovative methodology was developed aimed at automatic detection of stents in IVUS images; stents are the cylindrical and metallic devices introduced in blood vessels when they are blocked. Concerning IVUS images, doctors are able to see the pictures of arteriesʼ contours and stents produced by the computer application created by UB researchers.

This computer vision system provides objectivity (subjective factors such as tiredness disappear), precision (image information processing is as good as expertsʼ vision) and speed (the software works while the intervention is taking place, so it guides and enables to correct it if necessary). The software reduces the time that doctors spend in analysing images and the number of images that can be analysed increases. Consequently, it is possible to perform a more reliable estimation of lesions and a higher number of interventions. Nowadays, coronary artery diseases are the number one cause of death in developed countries. In Europe, it represents 21% of deaths. Moreover, 80% of coronary interventions require stent implantation. Therefore, it is absolutely relevant to have tools to analyse in a quick and robust way how to do it. Stent machine detection enables doctors to diagnose the result of the intervention in a quicker and more objective way. Patientsʼ risk to suffer future heart attacks is reduced. Boston Scientific commercialises its produces in about one hundred countries all over the world. The product that they sell most is the echograph iLab; UB researchers developed the software for this device. The collaboration with the firm and the group from the hospital has led the research group to three international patents and a publication on the most prestigious congress on medical image analysis, MICCAI 2013.