Emerging Ethical Issues for DataLakes

点击翻译按钮获取中文摘要

The proposed project is a partnership with the Innovate UK-funded PathLAKE Consortium, which is applying computational techniques to cancer diagnostics. PATHLAKE is one of 5 Centres of Excellence in medical imaging and digital pathology funded by Innovate UK. It is a partnership between university researchers, NHS trusts and commercial organizations. The University of Warwick is leading the computational arm of PathLAKE, working together with partners and experts at the lead partner University Hospitals Coventry and Warwickshire NHS Trust, Royal Philips and teaching hospitals and universities at Belfast, Oxford and Nottingham, focussing on breast, prostate, lung and colon cancers. The digital labs involved have pooled annotated whole slide images (WSIs) of tissue relevant to the targeted cancers and have welcomed applications from commercial digital science companies outside the consortium to train and test algorithms with its data. Ethical issues are raised by what is involved in stocking the "datalake" of WSIs. For example, where does the original tissue come from and under what informed consent regime if any? These issues illustrate the intersection in medical imagery research and development of data ethics, tissue ethics and general research ethics. Extracting data from the data-lake, especially where it leads to the development of for-profit apps using algorithms trained and tested on Pathlake WSIs, raises further issues in the ethics of public-private partnerships, specifically those anticipated by Codes developed for data-driven research by NHSX. There are also ethical and conceptual issues connected with whether data in the lake are truly anonymised or de-identified when returning clinical benefit from algorithms to tissue donors from the participating NHS trusts may depend on (selective and temporary) de-identification. Since the concept of personal data is tied to the inferability of identity, and since super computing power involved in computational pathology may allow for very subtle inferences far beyond human capacities, it is far from obvious whether WSI data is personal or not. A mapping and preliminary discussion of all of these issues will result from the project, and findings will be distributed to the Centres of Excellence, Innovate UK, and the UK Government Office for AI, as well as the Office of the Information Commissioner, and NHS Research Ethics Committees.