Self-Learning Systems

in the Healthcare System

Report by the Working Group
Health Care, Medical Technology, Care

Executive Summary
Whether in prevention, early diagnosis or selecting the ideal treatment,
Artificial Intelligence (AI) and Machine Learning (ML) could soon be
playing a big part in ensuring that people receive better and more per-
sonalised medical care. There is a whole variety of ways that self-learn-
ing systems could be put to use in the medical practices and hospitals of
the future. For instance, doctors could use AI systems across the board
to help them evaluate imaging procedures, thereby obtaining more ac-
curate diagnoses. By using networked data, self-learning systems could
derive proposals for suitable preventive approaches or treatments, thus
helping medics and patients make important decisions.

Collaboration between humans and machines also has a great deal of poten-
tial in the care sector. For example, AI-supported voice recognition systems
could help care workers with routine tasks such as completing documentation,
thus freeing them up to spend more time with people in their care. Assistance
robots and AI-based technologies such as exoskeletons, meanwhile, could in
future make it possible for people to live independently well into their old age.

It is not just the sick who can benefit from medical AI applications, as smart-
phone apps and wearables can enable people to record and evaluate their
own health data, whether they are ill or not. This gives people the opportunity
to make their day-to-day life more healthy or identify symptoms of illness early
on. AI offers huge potential, particularly when it comes to the prevention and
early detection of illnesses. The focus is shifting from cure to prevention.

The aim of AI applications across all healthcare sectors is to unburden medical

and care staff and give them the best possible support. All the technological
achievements are centred firmly on the benefits for patients and people in
need of care.
1
EXECUTIVE SUMMARY

To ensure self-learning systems in the healthcare system are reliable, safe and
will work entirely in the interest of people – and to harness their full economic
potential at the same time – certain prerequisites and parameters need to be
put in place. Representatives from science, business, hospitals, social businesses,
and patient representative organisations have come together in the working
group Health Care, Medical Technology, Care of Plattform Lernende Systeme
to discuss the opportunities and challenges of self-learning systems in health
care. Their work is funded for a further three years. As an initial interim out-
come, the working group is formulating recommendations for action within
the following fields:

Health data

„ Make the personal health data of insured persons, healthcare providers and
cost carriers in standard care continuously usable for machine learning.
„ Share data from all healthcare sectors.
„ Make existing stores of data visible, open them up, assess their usability
and network them. Examples include publicly recorded data from the
cancer and heart valve register, study results, hygiene data, etc.
„ Establish a representative, structured and controlled health database that
could be organised in a Centre for Digital Health Data or Digital Health
Institute, as is the case in neighbouring European countries.
„ Give German and European research institutions and companies equality of
access to anonymised health data to ensure individual companies cannot
establish monopolies.
„ Store and train health data in localised architectures as far as possible, so as
to improve security and data protection and thus guarantee the protection
requirements of healthy and sick people alike.
„ Develop options for a legal and technical framework for the voluntary and
secure donation of data (including the data custodian model).
„ Make compliance with the requirements of the General Data Protection
Regulation (GDPR) – which is strict by international standards – an advan-
tage in global competition.

Build up expertise in medical training and in care provision

„ Integrate AI into the training of medical and care personnel, such as

expanding study courses and training programmes on digital health.
„ Reinforce further training in AI and statistical processes for medical and
healthcare purposes so that personnel understand the individual applica-
tions and can put them to optimum use.
„ Train medics, nurses and care staff for their new roles regarding patients,
e.g. communicating AI-supported diagnoses and treatment recommenda-
tions.
„ Create information and training offerings for patients.

Bring innovations to patients

„ Pave the way for AI innovations in the healthcare system with outstanding
research and drive forward lighthouse projects and pioneering use cases.
Utilise structural and project funding instruments that extend beyond purely
basic research to roll out AI projects in medical practice.

2
EXECUTIVE SUMMARY

„ Specify a clear process for clinical studies to assess AI-based medical prod-
ucts. Legislators, notified bodies and the industry are to work together to
develop processes for clinical studies that reliably review the security and
benefits of AI applications.
„ Design a process for approving AI-based medical technology that takes
account of the specific challenges affecting a dynamic technology that
changes as it learns. (For accessing the European market producers of
medical devices must undergo a so-called conformity assessment procedure.
For the purposes of readability, we refer to this process as approval.)
„ Finance health care in an innovation-friendly way so that AI innovations can
be assessed and, if appropriate, adopted into standard care as quickly as
possible based on clear criteria.
„ Investment in AI and digitalisation must not mean funding cuts in other
important areas of the healthcare system.
„ Design a clear legal liability framework that gives medical and care person-
nel legal certainty.

Ethical issues

„ Guarantee the data sovereignty of all parties. Defined rules for anonymisa-
tion, pseudonymisation, the right not to know and an opt-out option all
play a key role.
„ Carefully weigh up medical benefits against the protection of personal
rights when establishing the legal framework.
„ Put in place the prerequisites to ensure the decisions made by AI-based
systems and the background assumptions can be traced by all users.
„ Reflect on and discuss the consequences of AI for society’s understanding
of health, illness and the concept of humanity.
„ Conduct a clear, honest and evidence-based discussion of the opportunities
and risks associated with AI in the healthcare system that gives all stake-
holders the scope to contribute their ideas, interests and concerns.
„ Work in ethics committees at institutional and national levels to develop
standards for AI-based human-machine interactions in the healthcare system.

Imprint
Editor: Lernende Systeme – Germany‘s Platform for Artificial Intelligence | Managing Office | c/o acatech | Karolinenplatz 4 |
D-80333 München | [email protected] | www.plattform-lernende-systeme.de |
Follow us on Twitter: @LernendeSysteme | Status: June 2019 | Image credit: vectorfusionart / Adobe Stock
This executive summary is based on: Plattform Lernende Systeme (Ed.): Lernende Systeme im
Gesundheits­wesen – Grundlagen, Szenarien und Gestaltungsoptionen (Bericht der AG Gesundheit,
Medizintechnik und Pflege), Munich 2019. The original version of this publication is available at:
https://www.plattform-lernende-systeme.de/publikationen.html