© National Cancer Institute / Unsplash

16 August 2019

Berlin‘s Charité works on using AI to improve the diagnosis and treatment of this widespread disease.

Whether high blood pressure, heart failure, coronary heart disease or cardiac arrhythmia: cardiovascular diseases have been the most frequent cause of death in Germany for decades. The good news: according to the German Heart Report 2019, fewer and fewer people are dying of cardiovascular diseases in Germany. In 2016, 2.1 per cent fewer patients died from cardiovascular diseases than in 2014, according to the report. Nevertheless, the German Society for Cardiology - Heart and Circulatory Research (DGK) is only cautious about rejoicing, as cardiovascular diseases remain by far the number one cause of death in Germany, accounting for 37.2 per cent of all deaths. Experts are also concerned about another development: in 2017 there were 1.71 million hospital admis-sions due to heart disease. This is more than 37,800 above the 2015 figure, a rise of 1.5 per cent. Increases are particularly noticeable in the case of heart valve diseases, cardiac arrhythmia and heart failure. For Prof. Dr. Hugo A. Katus, President of the DGK, "these are also the diseases which become more frequent with increasing age," in the light of the fact that a reversal of the trend cannot be expected in view of the ageing population in Germany.

Progress with treatment options

Despite this bad news, there are also positive things to report. It has after all been possible to reduce the mortality rate of heart failure patients by almost 11 per cent. "This is, of course, a great success that has not been achieved without reason," explains Katus. "On the one hand, we have made significant progress with treatment options in both the drug and interventional and technological areas. On the other hand, we have greater aware-ness among physicians of the fact that there are better treatment options available for heart failure and the associated arrhythmias. "Enormous progress has also been made in prevention and diagnosis: for example, ischaemic heart diseases, i.e. events triggered by circulatory disorders in the heart, have been reduced. "This shows us that our prevention strategies are beginning to take effect, and I find that very impressive," says the DGK President, at the same time wishing to see further targeted efforts to combat this fatal widespread disease.

Charité relies on artificial intelligence

Such a request is received with great interest at Charité Berlin. Cardiac surgery and cardiology at the Charité University Clinic and the German Heart Center Berlin are cardiovascular centres renowned worldwide with the departments of cardiology, pediatric cardiology and congenital heart defects, cardiac surgery and pediatric cardiac surgery. Charité carries out all cardiological interventions and cardiosurgical operations, and also offers the entire range of intensive care medicine. In addition to application in the clinic itself, various inter-disciplinary teams work on forward-looking methods to better predict the risk of heart disease and to be able to treat already affected patients more efficiently. The innovative approaches have one thing in common: they use artificial intelligence and machine learn-ing to analyse clinical information and image data, to recognize and interpret known patterns, and to apply them in the areas of prevention, diagnostics and therapy planning.

Prof. Dr. med. Marc Dewey, Charité Berlin © Wiebke Peitz

Automated diagnostics of the heart with neuronal networks

The developments of the AG Noninvasive Cardiovascular Imaging Heisenberg Research Group at the Institute of Radiology are exciting. "Radiology is an incredibly exciting subject - almost every patient treated at Charité receives radiological examina-tions," explains its director, Prof. Dr. med. Marc Dewey, "AI could help us to further optimise this frequent and important use of non-invasive imaging.” Such an improvement is particularly in demand in oncological and cardiovascular medicine. It is easy to explain why: every year, over two million cardiac catheter examinations are carried out in the EU without therapy. In Germany, more than half of the almost 900,000 interventions (German) carried out each year show that there is no narrowing of the coronary arteries and that therefore catheter treatment is not possible. This is problematic because outpatient exam-inations are not only exhausting for the patient, but above all costly. If some of these examinations could be replaced by computer tomography (CT), this would not only benefit the patient, but the financial savings would also be impressive. How high these could be is currently being investigated on 3,500 patients in 16 countries as part of the EU-funded DISCHARGE-Project. "The results are eagerly awaited," reports Marc Dewey. But even if they turn out to be promising, one problem still remains: "Although the cardiac CT is carried out in just a few seconds, an experienced radiologist needs about 30 minutes on average to analyse a patient's examination," says the head of the Institute of Radiology. He is convinced that the process can be accelerated and optimised if artificial intelligence and radiology go hand in hand. „Bionic Radiologist“ is the name of the pioneering concept, which has already been presented in medical journals. On the one hand, doctors can better decide which type of imaging examination the patient needs. On the other hand, automated analysis of radiological images can achieve greater consistency at any time of day and in any hospital. "And thirdly, therapy recommendations can be closely linked to the results of the examinations in order to make maximum use of therapy options and not overlook opportunities," expects Dewey.

Cardiac Risk App

The first steps in this direction have already been taken. "We have developed a prediction for the risk of suffering from coronary heart disease - the narrowing of coronary arteries," reports Marc Dewey, "this succeeded in an international consortium of more than 70 centres in more than 20 countries worldwide, headed by Charité.” The internationality is no coincidence: every day, around 50 million people worldwide complain of chest pain to their doctor. The assessment of the individual risk has so far been inaccurate and often leads to complications during treatment. Thanks to well validated patient and image data, the team has now been able to improve this risk assessment. "The data is now being used to make this prediction digitally widely available and to link it to the automatic analysis of CT image data of the heart," says Dewey. In a next step, these findings are to be combined in an app and made available to attending physicians, but this "must be thoroughly clinically tested, for example in the context of the new forms of care provided by the Innovation Fund (German), before it can be integrated into everyday life". Dewey and his team are not the only ones convinced that the Cardiac Risk App will be of great benefit to patients in hospitals. The development was backed as one of four innovations by the first „Digital Health Accelerator“ of Berlin Health Innovations, the joint technology transfer unit of the Berlin Institute of Health (BIH) and Charité.

Prof. Dr. med. Titus Kühne, Charité Berlin © Prof. Dr. med. Titus Kühne

Cardiology weather forecast  

"It's important that our research doesn't take place in an ivory tower. We are not working altruistically," emphasizes Dewey's colleague, Prof. Dr. med. Titus Kühne from the from the Institute for Cardiovascular Computer-Assisted Medicine at Charité. "We take care not to generate new questions with research, but to create new products and, in the next step, jobs that raise new research questions. Especially in the field of cardiology, there are enough applications and enough demand." Kühne and his interdisciplinary team of mathe-maticians, engineers and physicians at Charité are researching, among other things, how machine learning can remove disturbing artefacts such as pacemakers from magnetic resonance imaging (MRI) images. Or how only a partial picture can be taken and completed using artificial intelligence for a faster understanding of the patient. "But I personally see the greatest scope in the analysis of the image data," adds Kühne. "I compare it with weather forecasts. If I have different information from weather stations today, I can predict the weather for the next four to five days. We can then use the images in silico (note: by computer simulations) to test whether the therapy option is promising and how the treatment is going." This knowledge can in turn shorten operation times, avoid ineffective therapies and subsequent operations and thus save up to 30 per cent in health-care costs. Until now, patients with heart diseases have often been discharged from hospital prematurely because they had good values. In everyday life, on the other hand, they were exposed to other stressful situations, resulting in renewed heart problems.

ArtiCardio: individual operation and therapy planning

The fact that Kühne and his team's research does not take place in an ivory tower is demonstrated by a wide variety of successful collaborations beyond the boundaries of the institute, for instance collaborations with research associations such as the Fraunhofer Institute or interdisciplinary networking within the framework of the Berlin Centre for Machine Learning (BZML). Cooperation with industry is also crucial, as the „ArtiCardio“ project with Siemens Healthineers and the Berlin SME 1000shapes proves: the starting point was MRI images of children with a constriction of the main artery, the most common congenital heart defect, and with heart valve defects, the most commonly acquired heart diseases. Mathematical methods were used to calculate the flow conditions in the blood vessels. This enabled the researchers to determine how the patient’s blood pressure is in a relaxed state and how he reacts under stress, thus predicting the outcome of various surgical therapies. Prof. Kühne describes the procedure as follows: "Normally, I have to use large computer centres for such complex processes, and you can't afford that for the individual patient, because I want to treat several people. Through artificial intelligence, I can speed up processing so that I can interactively perceive changes in real time and make interventions. This acceleration is a big help for the doctor." And not only for him: if the procedure is integrated into daily clinical routine, every patient can benefit from improved treatment. According to Prof. Kühne, the licensing of the software is only a matter of time because: "The results are fantastic". They are optimistic that there will be even more good news to read in future German heart reports.