Radiology Reimagined: Advancing Clinical Practice Through AI Innovation (2025)-"Do the Right (AI) Thing: Ethical and Legal Aspects"

"Do the Right (AI) Thing: Ethical and Legal Aspects" – Dr. Charles Kahn

Pdf Summary

This presentation by Dr. Charles E. Kahn Jr. from the University of Pennsylvania explores the ethical principles guiding the use of artificial intelligence (AI) in radiology, anchored in biomedical ethics concepts: autonomy, beneficence, and justice. Autonomy emphasizes respecting patients’ rights to informed choice, beneficence focuses on acting in patients’ best interests and avoiding harm, while justice demands health equity and fair distribution of benefits and harms.<br /><br />Key regulatory frameworks such as the U.S. Government’s AI Bill of Rights require “notice and explanation” to patients about AI's role in their care to promote trust and informed consent, although disclosure should be balanced with practicality and risk of harm. Safety and effectiveness are overseen by entities like the U.S. FDA and European CE marking, with AI tools classified by risk and regulatory rigor.<br /><br />AI's performance is compared against human readers in tasks like mammography screening, showing promising accuracy but raising questions about whether AI must surpass all humans or serve as an aid. Ethical risks include reduced physician skills from overreliance, ignoring clinical context, false certainty from binary outputs, AI’s opaque “black box” nature, and inaccurate or biased decisions due to dataset limitations or technical bias.<br /><br />Data privacy is critical, involving careful de-identification of images and reports. AI hallucinations—fabricated responses from large language models lacking domain knowledge—highlight potential misinformation risks. Biases in AI—due to demographic, clinical, or automation bias—can perpetuate inequities, as demonstrated by algorithms correlating healthcare costs rather than illness severity, disadvantaging marginalized groups.<br /><br />Fairness metrics such as group fairness, demographic fairness, and equalized odds help evaluate AI equity. The future-ai.eu initiative and related guidance promote trustworthy AI emphasizing fairness, traceability, robustness, and explainability.<br /><br />Ultimately, ethical imperatives call for AI systems that promote patient well-being, minimize harm, uphold human dignity and privacy, and ensure just distribution of benefits, aligning technological advances with foundational medical ethics principles. Radiologists are encouraged to engage with RSNA resources for ongoing education in AI ethics.

Keywords

Artificial Intelligence in Radiology

Biomedical Ethics

Autonomy in Healthcare

Beneficence

Justice and Health Equity

AI Bill of Rights

FDA AI Regulation

AI Performance in Mammography

Ethical Risks of AI

Data Privacy and Bias in AI