Iterative reconstruction artefact removal using null-space networks

Casper O. da Costa-Luis, Andrew J. Reader

Track: short paper

Keywords: null space.neural network, data consistency, image reconstruction, PET

Abstract: Incorporation of resolution modelling (RM) into iterative reconstruction produces Gibbs ringing artefacts which adversely affect clinically used metrics such as SUV${}_{\mbox{max}}$. We propose the use of a null space network as a regulariser to compensate for these artefacts without introducing bias.

Paper Type: methodological development