Machine Learning in Cardiac CT Image Reconstruction

Beschreibung

This dissertation focuses on the removal of cardiac CT imaging artifacts caused by motion and metal implants. A combination of model-based data synthesis and subsequent data-driven learning of image enhancement methods is proposed. Forward models for virtual artifact generation are developed by incorporating prior knowledge about the cardiac anatomy and CT imaging physics. They form the counterpart of resulting learning-based backward models, which achieve significant reduction of artifacts during testing on real data.

Autor*in

Tanja Loßau

Themen in »Machine Learning in Cardiac CT Image Reconstruction«

Artifact Quantification Cardiac Computed Tomography Convolutional Neural Network Image Reconstruction Machine Learning Metal Artifact Removal Motion Compensation

Stimmen zu »Machine Learning in Cardiac CT Image Reconstruction«

Details

Link teilen