AI-based methods for image reconstruction and improvement are increasingly developed to be used in clinical routine. The objective of this session is to learn about the methods to train, test and validate these algorithms, to know about the benefits and potential problems when using them, and to understand how quality of the algorithms and images can be assessed and what is mandatory to achieve useful methods for clinical applications especially with regard to training, test and validation data.
Chairpersons' introduction
Elmar Kotter, Freiburg Im Breisgau / Germany
Christoph Hoeschen, Magdeburg / Germany
Specification of training, test, and validation data from a clinical perspective
Federica Vernuccio, Palermo / Italy
Problems of assessing AI-based CT image reconstruction, denoising or artefact reduction
Marc Kachelrieß, Heidelberg / Germany
Variable phantoms as a potential solution
Christoph Hoeschen, Magdeburg / Germany
Evaluation of image quality after AI-based image reconstruction
Mika Kortesniemi, Espoo / Finland
Data analysis and statistical tests as a potential assessment strategy
Sarina Thomas, Oslo / Norway
Panel discussion: What do you need for quality assessment when using AI image processing?