Sessions

Remarkable progress in artificial intelligence (AI) algorithms performance, and the fast growth in “real world” data available in electronic form generate high hopes for healthcare quality, efficiency, and accessibility improvements. But this game changing progress also causes growing concerns about the effects of growing AI use and its unintended, unanticipated, or even intentionally unethical consequences. Numerous issues and limitations of the algorithms and data used in healthcare and beyond have become more visible, and several organizations and researchers have proposed advice and guidelines to help address these concerns, issues, and limitations. Principles of Responsible AI are now promoted by several important organizations and stakeholders in the AI industry, but there is a need to move these principles towards practical realization and application in real-world scenarios. This panel will address several key aspects of responsible AI in health: explainability and interpretability; bias and fairness; reliability, reusability, and efficiency; privacy and confidentiality protection.

Patients with cancer starting invasive treatment programs often develop depression that physicians struggle to recognize at an early stage. We developed a prediction model for early identification of patients at risk for depression within the first month of chemo- or radiotherapy treatment to assist physicians and healthcare workers.