Ethics of AI in Breast Cancer Diagnostics

The goal of the research project from ENERI is to use a decentralized federated learning technique to incorporate artificial intelligence (AI) into medical diagnostics in hospitals throughout Europe, and to create an industrial prototype that will eventually lead to a commercial product. The technology improves the early diagnosis of breast cancer using MRI analysis by allowing hospitals to contribute to a centrally hosted AI model without sending raw patient data. In order to find predictive cancer traits, hospitals use machine learning algorithms to process local MRI data, updating the AI model in real time. Data safety is ensured by strong cybersecurity and encryption techniques. Doctors receive radiological output from the AI system, which helps with early diagnosis and therapy. Possible ethical issues include practitioner deskilling, dataset biases, accessibility and equity, AI explainability, and patient communication about AI-driven diagnosis.

Artificial intelligence in diagnosing medical issues such as breast cancer offers a chance for better clinical decision-making, increased efficiency, and improvement in early breast-cancer diagnosis. This new technology also brings up ethical questions about bias, professional obligations, honesty, accessibility, and the function of human supervision in healthcare.

Deskilling and the changing role of the clinician

The job of radiologists and other medical workers is changing as AI systems get more accurate and widely used. As raised during class discussion, radiologists and oncologists might become less central and mainly act as supervisors verifying results produced by AI. AI does not replace human evaluation, even if it can increase diagnostic efficiency and accuracy. The primary concern is that if AI is used too much, the physician's ability to diagnose problems may eventually deteriorate.

Whether AI should be considered an extra tool or a substitute for clinical decision-making was a topic of discussion. Even if a doctor misuses AI, they are still accountable for mistakes made. The use of AI as a justification for mistakes or negligence raises ethical questions. The scientific literature highlights that, although AI can offer insightful information, it does not have any of the clinical reasoning and contextual judgment that human physicians possess (4). Even as AI becomes more common, healthcare workers should continue to use their education to maintain their diagnostic skills in order to reduce these potential hazards.

Bias and representation in the data

Whether AI models accurately reflect a range of patient demographics is an important question. The class debate brought to light the fact that certain demographic groups can be underrepresented in training data, which could result in bias in AI-based diagnoses. Since AI can evaluate large amounts of data from many sources, some have claimed that it is by nature less biased than human research. Others argued that AI bias is a continuing problem, particularly when models are trained mostly on data from high-income or Western populations (2).

Getting rid of healthcare AI bias requires developers to make sure training datasets are diverse in order to develop fair AI systems. Research indicates that AI models trained on a variety of populations increase diagnostic accuracy for a range of socioeconomic and ethnic groups (3). To avoid inequalities in healthcare outcomes, underrepresented populations must be included in AI training datasets.

Explainability, accountability, and the "black box"

Another ethical question in AI decision-making is whether physicians are able to understand and communicate AI-generated diagnoses to patients. Informed consent and patient trust are compromised if a doctor is not able to understand an AI's decision-making process. Transparency and accountability are complicated by AI's "black box" nature, in which computers make judgments without providing a justification (5). Rather than accepting AI-generated results at face value, a doctor has to be knowledgeable enough to question them.

This raises the question of whether diagnostic errors are more likely to be made by artificial intelligence or by human doctors. Studies show that if AI is trained on large datasets, it can identify some cancers more correctly than human radiologists (1). However, mistakes still occur, particularly when AI lacks sufficient data. AI should therefore be utilized as an additional tool rather than as the sole decision-maker.

Accuracy over time

Another discussion topic was whether AI becomes more accurate with additional data over time. Machine learning models get more accurate and reliable as more cases are examined. However, in early deployment stages - when AI systems might not be completely accurate and mistake-proof - this technology still raises ethical questions. AI should be regularly reviewed and upgraded to address this and avoid dangerous misdiagnosis.

Conclusion

There are possible problems in incorporating AI into breast-cancer diagnoses while balancing ethical responsibilities and technological advancements. By promoting earlier and easier cancer identification, AI has the potential to improve patient outcomes. It is important to address concerns of accountability, transparency, and bias. In the end, doctors have an ethical duty to make sure AI is used properly and does not hinder patient care.

Although AI has promise in medical diagnosis, its application must be ethically justified through the measures described above. AI can help doctors, but it should not take the place of human knowledge. Integrating AI into healthcare ethically requires addressing bias, guaranteeing transparency, and making sure humans check the information presented by AI. To maintain a balance between innovation and patient-centered care, the technology must be continuously evaluated and adjusted.

References

1. McKinney, S. M., Sieniek, M., Godbole, V., Godwin, J., Antropova, N., Ashrafian, H., et al. International evaluation of an AI system for breast cancer screening. Nature. 2020;577(7788):89–94. https://www.nature.com/articles/s41586-019-1799-6
2. Buolamwini, J., & Gebru, T. Gender shades: intersectional accuracy disparities in commercial gender classification. Proc Mach Learn Res. 2018;81:1–15. http://proceedings.mlr.press/v81/buolamwini18a.html
3. Larrazabal, A. J., Nieto, N., Peterson, V., Milone, D. H., & Ferrante, E. Gender imbalance in medical imaging datasets produces biased classifiers for computer-aided diagnosis. Proc Natl Acad Sci U S A. 2020;117(23):12592–12594. https://www.pnas.org/content/117/23/12592
4. Hagendorff, T. The ethics of AI ethics: an evaluation of guidelines. Minds Mach (Dordr). 2020;30(1):99–120. https://link.springer.com/article/10.1007/s11023-020-09517-8
5. Miller, T. Explanation in artificial intelligence: insights from the social sciences. Artif Intell. 2019;267:1–38. https://www.sciencedirect.com/science/article/abs/pii/S0004370218305988
6. European Network of Research Ethics and Research Integrity (ENERI). AI and research ethics: challenges and considerations. 2025. https://classroom.eneri.eu/node/401