Go to JKU Homepage
Faculty of Medicine
What's that?

Institutes, schools, other departments, and programs create their own web content and menus.

To help you better navigate the site, see here where you are at the moment.

AI Helping to Predict Medical Treatment Complications
 

Cutting-edge technology: Artificial intelligence developed at RISC can predict complications when treating aortic aneurysms.

[Translate to Englisch:] Credit: RISC

Aortic aneurysms, a dangerous enlargement of the aorta which could result in aortic rupture, are among the most common vascular diseases, killing thousands each year in Europe and in the USA.  Open surgery puts a heavy strain on the body and requires a longer recovery time and while the so-called endovascular treatment is a minimally invasive procedure, it can result in a number of complications after surgery. Endovascular treatment involves placing an endo-vascular stent through small incisions at the top of each leg into the aorta, relieving the vessel and helping blood to flow, thus preventing any ruptures.

The EndoPredictor project involves researchers at RISC Software Ltd. as well as physicians at the Kepler University Hospital and at MATTES Medical Imaging Ltd. The project focuses on developing methods to extract and create digital abdominal aorta and aneurysm scans. Researchers used anonymous information provided by 50 patients consisting of computerized tomography angiography scans (CT-A) taken before the procedure and then at several follow-up appointments after endovascular treatment. The scans were used to model aortic and stent-grafts simulating blood flow through the stent-graph and calculating any potential changes in the stent-graft’s position and shape during follow-up appointments.

Researchers studied 201 CT-A scan sets and calculated 42 measurements in each case. These measurements describe the aorta and stent-graft’s shape as well as features derived from the simulated blood flow. Researchers studied which measures show statistical correlations with complications such as leaks, vasoconstriction or vascular occlusions. Predicting these complications was made using a specially developed procedure. The result is a software system for the automated prediction of post-operative complications after endovascular treatments. The prediction rate of accuracy was up to 88%.

The prediction method was developed based on machine learning and not only automatically detects corresponding characteristics in the data, but also learns from new data sets. A new method to validate the selection of characteristics was implemented along with methods to simultaneously train the prediction system.

A patent for the procedure is currently pending. The project findings will subsequently be scientifically published and the process will continue to advance and to eventually become a finished product.