An efficient semi-supervised quality control system trained using physics-based MRI-artefact generators and adversarial training

Daniele Ravi, Frederik Barkhof, Daniel C. Alexander, Lemuel Puglisi, Geoffrey J.M. Parker, Arman Eshaghi

Abstract: Highlights•We have developed a semi-supervised approach to identify artefacts in brain MRI.•Our solution utilizes physics-based artefact generators, multi-domain features, and artefact-specific selection.•We employ the physics-based artefact generators to create synthetic brain MRI scans.•The proposed pipeline can be used for monitoring MRI scan quality.

External IDs:doi:10.1016/j.media.2023.103033