Artificial intelligence-based self-learning in medical imaging

Abstract

For machine learning for a medical imager, results created for individual patients are used to generate the ground truth. The acceptance or change for examining an individual patient is used as the ground truth instead of using a further expert study for the purposes of machine training. In this way, the medical imager creates both samples and ground truth as part of every-day, on-going examinations of patients in the production environment. Machine training is performed based on these samples, and the machine-learned network may then be applied for imaging further patients. For example, the medical imager self-optimizes or self-learns, allowing for updating the machine-learned network more rapidly (e.g., keeping pace with changes in practice in a lower cost and less time-consuming approach and/or updating localized to a practice) in the production environment.

Dependencies are inferred from the claim text.