Learning coalitions: how AI and experts can improve their collective decisions from images

Abstract

Deep learning technologies have enabled the development of artificial intelligence systems for the recognition of anomalies in images or for predicting possible effects on the imaged object. For the application to medicine, we propose a new approach to learning and continuous improvement of models, based on a coalition of heterogeneous hospitals sharing the same care methods and continuously improving decisions in a common way to achieve the same quality of clinical decisions for all. The models developed and the humans constitute a learning coalition and the progresses in the quality of the decisions made by the models and by the human experts are quantified in the respect of ethical rules adopted by the coalition.

Ähnliche Arbeiten