Developing digital solutions following the ethical principles in the healthcare sector is a responsibility AI4LUNGS Project takes seriously. Patients and their families, especially in moments of vulnerability, deserve the utmost protection when it comes to data usage and algorithmic decision-making. This is why the AI4LUNGS project is committed to creating AI-based stratification tools that follow the highest ethical and regulatory standards.
In this blog, we are looking at the seven ethical principles elaborated by the project partners Marcel Rebbert and Isabelle Schlegel from Deloitte, and Magdalena Kogut from TLX.
Upholding Ethical Standards in AI Development
The AI4LUNGS project recognises compliance with established legal frameworks like the General Data Protection Regulation (GDPR) and forthcoming regulations such as the European AI Act. To ensure ethical excellence, the project has implemented the AI4LUNGS Personal Data Protection Manual and Regulatory Acceptability Plan and follows core digital ethics principles inspired by European Union values.
Our ethical guidelines align with the Ethics Guidelines for Trustworthy AI by the EU High-Level Expert Group on AI. Additionally, we have integrated recommendations from organisations like UNESCO, OECD, and the G7, ensuring our project is both ethically and contextually tailored to the healthcare sector.
The Seven Principles AI4LUNGS Commits
Transparency
"We plan, develop, test, and implement our solutions transparently."
AI4LUNGS strives to provide a clear understanding of our solutions, detailing their development, integration into clinical workflows, intended benefits and potential risks. When using AI models, we follow an explanatory approach; AI models should be explainable to the extent possible and according to the capacity of the relevant audience, whether physicians, patients, or others.
Diversity, Non-discrimination and Fairness
"We take measures to minimise bias in our solutions."
AI4LUNGS involves a diverse range of stakeholders in the development process to ensure multiple perspectives are considered in data use and technology development. Our user-centred design prioritises equal access and usability for all, regardless of age, gender, ethnicity or other factors. We use representative data to develop our solutions, aiming to avoid bias and discrimination.
Technical Robustness and Safety
"We ensure that our AI models generate consistent results and are safe to use in clinical settings."
AI4LUNGS’ AI models are developed to account for variations in clinical performance, ensuring they perform reliably under different conditions. We continuously track and assess the efficiency of our AI models to confirm they operate as intended and to evaluate if they impose any negative effects on individual patients and doctors. In addition, our solutions are intended to only be used by appropriately trained healthcare professionals to prevent any risk of harm to patients.
Accountability
"We acknowledge our responsibility for the development of trustworthy healthcare AI models and solutions."
AI4LUNGS have built an effective governance with a clear reporting structure to represent patients’ interests responsibly. We have appropriate measures in place if issues arise with our AI systems. This includes the ability to submit complaints and provide redress for patients and healthcare professionals. Our accountability also includes maintaining auditability, minimising harm, reporting negative impact, and communicating research-related limitations to the public.
Societal and Environmental Well-being
"We promote the societal benefit as well as a sustainable use of AI models."
Human well-being has always been at the heart of healthcare. Our AI models and solutions are designed to support human well-being, improve patient outcomes, facilitate healthcare professionals' work, and ensure affordable access. We also focus on interoperability and sustainable data processing to help contribute to the sustainable development of society.
Privacy and Data Governance
"We respect individuals’ privacy and protect personal data through a comprehensive data governance."
AI4LUNGS comply with applicable data protection regulations and implement privacy-preserving measures, such as pseudonymisation, anonymisation, and encryption. We have robust cybersecurity measures in place to protect our AI models to prevent unauthorised access, modification, and loss of personal data. Where required, we obtain the necessary permissions and inform patients and healthcare professionals about how their data will be used.
Human Agency and Oversight
"We protect human autonomy and the decision-making powers of healthcare professionals and patients."
AI4LUNGS’ AI models and solutions assist healthcare professionals in making informed medical decisions without transferring decision-making powers. The final care decisions remain a discussion between the healthcare professional and the patient. We ensure an appropriate level of human control and oversight and promote the human interpretability of our AI model outputs. Established protocols are in place to address situations where AI solutions produce low-confidence or conflicting results.
Ethical AI for Better Healthcare: AI4LUNGS’ Approach
By embedding ethics into every aspect of development, we reaffirm our commitment to creating responsible and impactful AI solutions that prioritise trust, transparency, and societal well-being.
Stay updated on the AI4LUNGS research development and progress by following the project on X and LinkedIn, and subscribing to our newsletter for the latest insights and advancements.