FANTAISIE

AI approaches such as tomographic image reconstruction and pre- or postprocessing techniques might offer better medical imaging such as CT, interventional imaging, PET and SPECT using potentially less ionising radiation exposure to patients. AI-processed images look always as being of high fidelity, making it difficult to identify missing information or introduced artefacts. Thus, potentially, misdiagnosis might happen. Incorrect diagnosis based on CT, PET or SPECT medical imaging procedures are a radiation protection issue as ionising radiation is applied to the patient and no benefit is generated. Justification as one key principle of medical application of ionising radiation is not given.

Due to the images suggesting high fidelity it is difficult to compare the quality of e.g. different CTs. Quality assurance is important as problems with the system hard- or software will be hardly detectable. This quality assurance will be difficult for AI processed CT as standard test objects can be detected as such by the software and presented always ideally covering occurring problems.

The aim is to enable quality assessment of CT imaging procedures which are AI processed CT imaging and to identify needs for regulation for this. Variable phantoms for chest CT imaging as software and hardware phantoms representing real anatomical structures with variable parts should be developed to generate test objects for which the ground truth is known, but where the system cannot easily identify it as a known test object. This will allow CT quality assessment and assurance and thus the safe use of improved imaging technologies.