14 January 2019

Dr. Dietmar Frey from Charité Berlin about the AI revolution in stroke diagnosis.

Figures demonstrate it in an unambiguous way. Each year 1.2 million people fall ill in Europe. In Germany there are 260,000 and about 15 per cent of these have yet another attack within a year. We are talking about a stroke. Although the clinical symptoms and the causes of a stroke of the cerebral vessels differ from one patient to the next, the clinical care does currently not have individual and personalised diagnosis- and therapy strategies.

For a number of years, Dr. Dietmar Frey and his interdisciplinary team of physicians, software developers and computer scientists has been working to find an alternative. Apart from being head of the research group “Predictive Modelling in Medicine“ at the Charité Berlin, the neuro-scientist signs responsible for the PREDICTioN 2020 project. An innovative simulation model was developed with the assistance of artificial intelligence that can provide a predictive diagnosis in stroke treatment by means of medical imaging.

^{Dr. Dietmar Frey, Charité Berlin © Dr. Dietmar Frey}

Diagnostic investigation and therapy on their way into the 21^st century

“It was our objective to use all digital data available of the patient. Based on our expertise with strokes we started first of all with the secondary prevention of stroke patients in the Charité”, describes Dr. Frey the beginnings of the project. “Contrary to the classic currently used clinic decisions that are based on population-based general values, our service allows for decisions and interventions that are based on quantifiable personalised risk of the relevant patient.”

In order to enable the treating physician to set up the best possible therapy for each patient, the team fell back on the so-called predictive “Hybrid Modelling-Technology” - that is linking bio-physical simulations of biological parameters in combination with machine learning algorithms.

It is a long path with many obstacles to be overcome - from clinical studies via developing a product / market fits right up to the certification and approval as medical product - before a prototype - as in the case of PREDICTioN 2020 - can be introduced successfully to the market.

Innovation by keeping an eye on the Data Protection

Where Dr. Frey is concerned, the location of Berlin fulfils to a large part the requirements he needs to further the development of technology in the section of artificial intelligence. It is time to take the opportunities offered and to set the course for the future: “There are outstanding scientific institutions available in Berlin, where experts are being trained. And there is an eco-system due to the good linking of scientific institutions.” In order to catch up with the market leaders in the U.S. and Great Britain, and despite successfully commercialising the field and excellent prerequisites, it lacks investors ready to take a risk to promote the section of Digital Health decisively with their investments.

At least in the European community, cards are being reshuffled for a researcher in respect of the sensitive subject of data and data protection. He understands the new European ‘Datenschutz-Grundverordnung (DSGVO)‘ [European Data Protection Regulation] to be a great opportunity to build up “an ethical data management adapted to patients“ as it is indispensable for AI-based models to have access to and to analyse longitudinal data that can demonstrate individual processes of change. “These are very difficult to come by also for owed to the fact that within institutes (above all in stationary areas / hospitals) data are not collected on principle on long-term processes and outcome results. It is at this point that politics could assist decisively by obliging the health system to collect outcome-based longitudinal data for every patient.“

Frey, however, does not call for collecting data in huge uncontrollable quantities but stands for a reasonable target-oriented usage by focussing on autonomy, transparency, data protection and securing data: “The most important challenge is to use data in order to change things for the better of people but focussing on patients and people most of all. The patients must remain owner of his data.” The challenge for research, ethics and politics will be, however, how to harmonise the aspects of data storage and data control by taking the public welfare into consideration. The neuro scientist warns that this development may lead to ‘making data available compulsorily‘.

Further Development and Adaption

Working with a visionary technology such as Artificial Intelligence that develops with such a fast pace induces the wish to look into the future. The team of PREDICTioN 2020 is planning to create improved possibilities for a primary prevention of strokes, based on their current model. ”On the one hand physicians are given a ‘Monitoring and Risk Stratification Tool‘ and on the other hand the patient is put in a position to reduce risks by changing his attitudes“, explains Dr. Frey, who is thinking further ahead and has already identified potentially adaptive scenarios for his model: ”Should we be able to successfully demonstrate that by using digital data based on AI models, we are in a position to improve the patients’ care in the area of strokes significantly, this technology could be applied in principle to other chronic illnesses (diabetes, hypertension, etc.).“ It seems that the potential of Artificial Intelligence in the Digital Health Sector has not been fully exploited by a long run.