ESR seeks changes to Europe's AI white paper

2020 06 22 16 26 9539 Artificial Intelligence Face 400

The European Society of Radiology (ESR) has welcomed the European Commission's recent white paper on artificial intelligence (AI), but it is calling on European institutions to consider a number of suggestions aimed at facilitating the safe development and adoption of AI technology in healthcare.

Specifically, the ESR has recommendations in the areas of data accessibility, transparency, data privacy, accountability, training and skills, telemedicine, and harmonization at the European Union (EU) level.

Facilitating access to data

Lack of access to large amounts of varied datasets from across EU member states hampers the ability of healthcare actors to optimize their services, as well as the potential for researchers to achieve breakthroughs, according to the ESR. As a result, the society supports the implementation of an EU Health Data Space to facilitate cooperation across the EU.

"A strong, efficient, and interoperable European framework for the collection of data across borders would foster research development, early detection, and early diagnoses of diseases," the ESR noted. "These outcomes can only be achieved by establishing a robust governance model that sets out how to use health data, while respecting the individual's privacy and transparency rights. For full efficiency and universal use, the ESR believes in drafting EU guidelines for health data to be in the same machine-readable format irrespective of its origin."

Providing data use transparency

Patients should maintain control over their personal information, according to the ESR.

"Besides, the ESR trusts that an individual informed on the responsible management of his or her data and its potential for patient care improvement, will be inclined to consent to the processing of it," it stated. "The European Commission should support initiatives, such as recommendations and codes of conduct, that are aimed at guiding healthcare professionals and academia in safely collecting, processing, using, and reusing data in clinical practice and research."

Furthermore, the ESR believes that data collectors for AI algorithms should provide clear and openly accessible information about how the data is being used.

Ensuring privacy

The ESR emphasized that health data collected for public benefit must fully respect individuals' privacy. Data anonymization or pseudoanonymization should be utilized to protect the privacy of patients in accordance with the General Data Protection Regulation (GDPR).

The society believes that effective application of anonymization would foster cross-border and cross-sector exchanges as well as effective research.

"Indeed, anonymized data under the regulation on the free flow of nonpersonal data across Member States, enables for data sharing that is exempt from full documentation or owner approval," the authors wrote.

Protecting patients, healthcare workers

As part of its support for data privacy and patient safety, the ESR also recommends an EU-level mechanism and an update of the Product Liability Directive for patients and healthcare workers to be protected and covered from any harm caused by breach of the law or the use of AI-operated medical devices.

"EU sectoral legislation (cfr. Medical Device Regulation in healthcare) should be adequately implemented and regularly assessed, considering latest technological developments," the society wrote. "Besides, a risk-based approach should be pursued in any cross-sectoral legislative and policy framework to simultaneously strengthen consumer protection, uphold legal certainty, and create a level playing field that allows innovation to excel in Europe."

Expanding training in AI

Adequate training for healthcare professionals and citizens is needed to unleash the full potential of AI in healthcare, according to the ESR. Consequently, the society is calling on the European Commission to further its measures laid out in its 2018 Digital Education Plan and push for jointly agreed EU quality indicators for continued medical education and also changes to university syllabi to include AI.

The ESR said it also welcomes the European Commission's efforts to close the digital skills gap with EU funding.

"Adequate investments through programmes including the Digital Europe Programme, Horizon Europe and ESF+ will help forge 'AI and digital-educated' organisations and citizens, a prerequisite to unleash the full potential of Artificial Intelligence society-wide and across sectors," they wrote.

Implementing telemedicine framework

Teleradiology can minimize the risk of COVID-19 infection for imaging professionals and also bridge inequality gaps in healthcare. But it can also potentially threaten the quality of care to patients, as well as radiologists' interactions with clinical colleagues.

With those considerations in mind, the ESR supports the implementation of an EU-level framework for telemedicine that would ensure high reporting standards, respect of the EU Health and Safety legislation, clear liability rules including an update of the product liability directive, verified patient consent and confidentiality, and full recognition of teleradiology as a medical act.

"Uniform standards for telemedicine would benefit clinicians, patients, and healthcare systems in their search for better access, quality, and efficiency of healthcare," the authors wrote. "Currently discussed legislative proposals, such as the digital services act and the update of the product liability directive, offer a window of opportunity to simplify and regulate the safe implementation of telemedicine in Member States."

Avoiding fragmentation

The ESR said that a fragmented scheme with various national schemes would hamper the full potential of AI to improve lives.

"The healthcare sector's research excellence and efficiency in the uptake of AI relies upon EU-wide cooperation, a framework for best practices exchanges between national initiatives and infrastructures as well as an efficient EU health data space," the ESR wrote. "Any initiative, from legislation for high-risk areas to nonbinding recommendations for the self-labeling of AI applications, should be furthered with the goal to build confidence among users and support the uptake of trustworthy AI across the EU."

The full ESR response can be downloaded on the society's website.

Page 1 of 109
Next Page