RACOON-AI Brain Tumor

RACOON-AI Brain Tumor is developing AI-supported analysis methods based on multi-parametric, quantitative MRI to improve diagnostics, therapy planning and follow-up of brain tumours. A nationwide standardised, multi-centre data set is being set up for this purpose.

The most important goals are

• Reliability of the method: Proof that quantitative MRI parameters provide comparable results at different locations.
• Non-invasive tumour characterisation: Prediction of molecular and epigenetic tumour characteristics directly from imaging.
• Prognosis estimation: Development of models to predict progression-free and overall survival.
• More precise image interpretation: Differentiation of active tumour parts, infiltration and oedema - if possible without contrast agent.
• Therapy monitoring: Differentiation between genuine tumour progression and therapy-related changes.
• In the long term, this will create a unique reference database that will set new standards for AI-based neuro-oncology.

A total of 621 patients (500 initial diagnoses of diffuse gliomas of the adult type and 121 glioma patients with the question of progress vs. pseudoprogress) are to be included.

RACOON-AI Brain Tumor is the first project to prospectively introduce a standardised, multi-parametric qMRI protocol at twelve university medical sites. Different scanner platforms, software versions and local IT infrastructures have to be harmonised in such a way that comparable quantitative parameters are produced. Several years of preparatory work have already been completed. Manufacturer-independent reproducibility is systematically checked with the help of phantom and travelling head examinations.

At the same time, a fully automated processing chain needs to be set up - from de-identification and upload to centralised post-processing in the containerised Kaapana/JIP environment. This requires robust, standardised workflows, high data quality and reliable technical integration of all locations.