Deep Dive #3: AI and health in Berlin

Hollywood and science fiction have shown us how humans and artificial intelligence can live together in the future. In the meantime, reality has almost caught up with fiction. And it turns out that in reality man and machine often enter into a fruitful partnership. So there is no reason to fear. Because in fact AI has nothing but friendly intentions. As it is so optimized by man, it is eager to support him to the best of its artificial knowledge and conscience.

AI and man - a cooperative mentor-pupil relationship

The areas of application of AI extend into just about every sector. The actual task of artificial intelligence is the real-time analysis and structuring of data. And this in a quantity which can no longer be handled by humans. It can be both image and text files. So far, AI systems have not been capable of independent interpretation com-pared to humans. They need mentoring and must be fed with so-called training data. This means that an extremely human teacher has to teach them what can be seen in the corresponding pictures and which parts of a text are relevant. On this basis, the AI system is able to recognize certain patterns and structures. A human component is always necessary. This is why it is usually called a weak AI. However, one should not be misled by this name. Also behind this AI there is a lot of potential.

AI systems can save lives

The healthcare industry is exploiting this potential and benefiting from artificial intelligence: after all, AI systems are precise and efficient helpers in the diagnosis, prevention and treatment of diseases. However, doctors still have the stethoscope firmly in their hands. As with other weak AIs, a kind of teacher, trainer or mentor must for example feed images and the corresponding, already known diagnoses into the system. This serves as a foundation for all further diagnoses. Thus, thanks to imported photographs, skin diseases such as skin cancer can be identified more quickly and accurately.

Furthermore, thanks to the decoding of the genome, humans are faced with a large amount of data. This is a quantity of data that AIs can analyse and structure much more easily in order to make faster progress towards curing cancer in the future. In the case of cancer, the most important thing is that the earlier the disease is detected the greater are the chances of a cure. For example, AI can detect changes in a DNA sequence which indicate a disease and use this information to tailor an appropriate therapy more precisely to an individual patient. Details and irregularities in the heart rate, which doctors cannot detect, ultimately lead to more accurate diagnoses of heart diseases.

However, AI systems are not only usable in the biological-scientific and diagnostic field. The administrative branch of the health system can also benefit from AIs. Systems can learn from the processes in hospitals and optimise them to increase efficiency. In digitised operating theatres, the patient is checked in digitally and the electronic checklist is processed digitally to ensure that no work step is overlooked. Before the operation, the patient's image data is broken down into individual segments. To ensure that surgeons make the right incisions during the operation, AI helps with orientation using a kind of navigation system. This possibility of orientation can help with minimally invasive surgical incisions to protect the patient as far as possible.

Berlin's exciting start-up and research landscape demands AI Health

Many exciting advances in the field of AI are taking place in the capital Berlin. First of all, the region has great potential as a research location: 50 to 65 professors and a wide variety of research institutions are working on the topic and implementing corresponding projects. In the health sector, the following courses of study and institutes can be found:

• Digital Health at Hasso Plattner Institute of the University of Potsdam: The main focus of the study programme for computer science and medical students is data protection, personalization and individual therapy options
• Center for Biomedical Image and Information Processing (HTW Berlin 2018) (HTW Berlin 2018): Focus on method development with image and signal data, transparency in data analysis, real-time data analysis and development of secure data infrastructure
• Special Research Area (SFB) 1294 Data Assimilation (University of Potsdam 2018): Erforschung von Anwendungsfeldern, unter anderem in der Biophysik und Medizin 

However, a large part of the growth in the AI sector is driven by start-ups. Throughout Germany there were 139 start-ups between 2012 and 2017, 48 per cent of which were in the Berlin-Brandenburg region. Of course, many of these were in the health¬care sector. By comparison, exciting examples do not have to hide behind national or international AI colleagues and can make the everyday healthcare of patients and doctors much easier.

https://www.uni-potsdam.de/de/sfb1294.htmlHere you can access the study "Artificial Intelligence in Berlin and Brandenburg".

After seven years of development, the Ada Health AI app was launched for the first time at the end of 2016. Just one year later, it shot to first place among the health apps and is already demonstrating what artificial intelligence can make possible in healthcare: the app creates an initial diagnosis and a practical overview of one's own health from the entered data or symptoms of illness. Diseases can already be detected and treated in their early stages. To ensure an accurate result, the app includes thousands of disease patterns. At the same time, it shows the app user and the physician the diagnostic path for maximum transparency.

The Audatic AI system, on the other hand, tries to support the deaf and hard of hearing. It detects annoying and disturbing background noise and can filter it out for the hearing aid. The hearing quality and thus the quality of life of users and affected persons can be increased enormously in this way.

LexaTexer collects a huge amount of information or data consisting of, for example, doctor's visits, prescriptions and doctor's notes. Basically, all information is included which leads to an error-free diagnosis and an improved quality of care for patients. In turn, fewer errors result in high cost savings.

The Boost Thyroidapp is specifically designed for thyroid gland patients who want to check their symptoms. Users feed the application with their weight, disease symptoms, information on medication and supplements. Another practical feature is a reminder to take medication. Patients are also kept up to date with the latest scientific findings on thyroid diseases.

xbird also takes advantage of large amounts of data and analysis by artificial intelligence to avert preventable diseases. Millions of data from smartphones, wearables and other smart user items are used to detect patterns for upcoming critical health events at an early stage. Medical professionals can in turn use this information to make better decisions about treatments. This technique is already being used in diabetes to prevent hypoglycemia or hyperglycemia.

To find out which innovations in the field of digital health are being developed at the Berlin location, the Senate Department for Economics, Energy and Public Enterprises announced the Digital Health Award. Among the more than 30 submissions, solutions that make use of AI repeatedly emerged. The companies Ada Health and Turbine have for example been nominated for the award.

You will find more information about the Digital Health Award here (German only).

These are just some of the many exciting new companies and innovations that can help to relieve the burden on the health system and, at best, leave waiting rooms empty. Even patients who do not have a doctor in the immediate vicinity can receive first aid and diagnosis. And Berlin-Brandenburg, with its progressive innovative spirit and unique start-up infrastructure, offers the ideal breeding ground for research, further development and use of artificial intelligence.

Where a large amount of data is involved, data protection naturally plays an impor-tant role. It is precisely this issue which still poses a major challenge for the AI sector. After all, sensitive patient data must not be disclosed to unauthorised third parties. Although data protection is still relatively simple in the case of X-rays, the protection of detailed diagnostic reports is much more complex and correspondingly more costly. In addition, the question also arises in this context as to whether a purely governmental or semi-governmental body should be responsible for data management. These still unanswered questions are in turn countered by numerous obvious and undeniable advantages of a far-reaching neural network.

With the "Deep Dive" series Projekt Zukunft regularly gives an insight into current technologies in the digital, media and creative industries and provides information about actors, trends and applications from Berlin.