Inter-modal selective 3D coding of PET-CT datasets

In this work we introduce a new selective coding approach suitable for co-registered multi-modal medical images and we apply it to large PET-CT volumes. Salience information guiding a space variant reconstruction quality of the anatomical volume (CT) is generated through an automatic analysis of the functional volume (PET). This allows a versatile multiple volume-of-interest coding with arbitrary 3D-shape and scaling-factors and without the need of side-information to be transmitted. The proposed solutions are suitable for critical applications where high and optimized compression ratio, minimization of human intervention and full diagnostic quality preservation are all required.