Neural Expert Systems in Medical Image Interpretation: Development, Use, and Ethical Issues

Abstract

In this paper, the appropriate level and role of neural network-based methodologies in the development and use of expert systems for medical image interpretation is investigated as technical, organizational, and social issues become intertwined. The notion of the information life cycle is applied to highlight ethical issues during the acquisition, processing and storage, dissemination and use of clinical information. These issues are further analyzed from a stakeholder perspective to accentuate the role of human agents in avoiding ethical risks. Relevant stakeholders, other than the key participants-namely system developers and medical users-are identified. The results of this analysis indicate that each stage of the development and use of a neural expert system entails ethical issues. Significantly, the responsibility for medical image interpretation is affected by contextual factors and should be shared amongst the main stakeholders. These conclusions are useful for the stakeholder groups that are conscious of their obligation to behave ethically and for researchers who wish to investigate further the ethical implications of artificial intelligence use in medicine. +

Clinical AI: opacity, accountability, responsibility and liability

“…Delvaux (2017) asserts that an AIS should assist the clinician; that the planning and final decision for the execution of a treatment must be made by a clinician. Pouloudi and Magoulas (2000) warn that an AIS's user is responsible for evaluating its outputs before using them. Whitby's (2015) insists that clinicians should maintain responsibility for outcomes when they use AISs and that clinicians ought not be allowed to escape that responsibility by blaming the AIS should negative outcomes arise.…”