Stents are tiny meshed wired tubes which function as scaffolds to re-open blocked blood vessels narrowed by disease. Artery disease remains the single largest killer in this world. Statistically every 4th person will die from its consequences with a grim future outlook on improvement. Stents are usually very effective and are often the most preferred treatment method for later stage disease.

More than 2 million stents are implanted every year worldwide, yet nearly 10% of them still fail. Think about it, that’s more than 200,000 affected patients every year. Additional procedures, expensive drug treatment and severely limited enjoyment of life are just a few of the consequences.

Our team made it their goal to understand the reasons WHY stents fail, how we can improve them and help more patients effectively.

The largest vessel bifurcation is the left main. It is commonly affected by disease and thus focus of many discussions amongst experts regarding the best treatment approach. Understanding difference in left mains in a large population can significantly help to improve treatment and translate it from a generic to a patient cantered approach, whereby individual differences are accounted for.

The coronary arteries consist of a left and a right main branch. Each continue to branch of into smaller and a smaller vessels. Bifurcations are the regions called where the vessels split.

The analysis of more than 200 patient’s medical images found that there are profound differences in the different bifurcations in one patient, but also vary significantly across patients with some common trends.  This has a significant impact in treating these vessel segments when diseased, whereby one stent may not fit all bifurcations or patients.

Advances in computing hardware and machine learning are revolutionising data intensive research. By accumulating a large database of coronary arteries, we have a unique opportunity to use machine learning to improve risk modelling and the prediction of heart disease. In particular, we are interested in how the latest approaches in artificial neural networks could be used to develop software capable of helping medical specialists with early detection of at-risk patients. Our ambition is to use medical image data together with fluid-flow simulations to train deep neural networks capable of identifying adverse geometries and advance segmentation techniques.