Genetics of deep-learning-derived neuroimaging endophenotypes for Alzheimer's Disease (Parent grant)

点击翻译按钮获取中文摘要

Supplement to Genetics of deep-learning-derived neuroimaging endophenotypes for Alzheimer’s Disease. Abstract AI/ML provides unprecedented opportunities for biomedical researchers, such as the quick identification of the genetic basis of diseases, including Alzheimer’s Disease (AD). For instance, the parent grant proposes new deep learning based approaches for deriving AD-relevant endophenotypes from neuroimaging data, and associating these endophenotypes to genetic data. It expects to discover new genes relevant to AD which may lead to a better understanding of the molecular basis of AD and potential new treatments. However, AI/ML methods could bring potential biases in the design and implementation of data collection, training data, as well as algorithm development. Such biases may lead to problematic findings and may further contribute to health disparity. Recent years have witnessed the heightened scholarly and societal discussion of principles of ethical AI; however, there is limited empirical data or evidence-based mechanisms that have demonstrated researchers’ knowledge, attitudes, or perspectives on ethical issues that impact the development of AI/ML algorithms or how they consider integrating research ethics into their work. Furthermore, how to develop and deliver effective AI ethics education is another issue that requires systematic scientific inquiry. This proposed supplement brings together AI researchers and bioethicists to create the first measure scale to measure medical AI researchers’ attitudes toward AI research principles (beneficence, non-maleficence, justice, and responsibility) and their knowledge about how to use these principles to guide ethical decision making in conducting Alzheimer’s Disease Research using AI through the use of case study vignettes. To create effective AI ethics education geared toward AI AD researchers, we bring in virtual-reality serious game designers to develop a VR-based, interactive application for education on ethical decision-making medical AI in research. Such an interactive and immersive mode of delivering educational materials has been shown to lead to more engagement, enjoyment, and higher effectiveness, compared to traditional educational channels. Information collected from researchers as well as a community advisory board will also inform the development of this AI ethics training program. The usability and effectiveness of the VR application will be evaluated using post-test survey and focus group.