An AI-driven framework for automated detection of giant cell arteritis in MRI

• Applicants: Professor Dr. Thorsten Bley; Privatdozent Dr. Julius Heidenreich; Professor Dr. Tobias Wech
• Subject Area: Rheumatology ; Radiology
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 568432998

Giant cell arteritis (GCA) is a systemic vasculitis characterized by inflammation of medium and large vessels. It notably affects the aorta and its branches, but especially the small superficial extracranial arteries as the superficial temporal and ophthalmic arteries. The autoimmune inflammation leads to vessel wall swelling and compromised blood flow, potentially causing severe complications such as vision loss, stroke, aortic aneurysms, or dissection if untreated. Current GCA diagnosis involves clinical assessment, laboratory tests, imaging, and sometimes histopathological confirmation. Temporal artery ultrasound is common but operator-dependent and often fails to consistently visualize all cranial vessel segments. Magnetic resonance imaging offers comprehensive imaging, capable of detecting segmental inflammations that ultrasound might miss. Modern MRI techniques acquire images of the superficial cranial arteries with high resolution and can reliably reveal vasculitis signs such as mural and perivascular enhancement. Thereby, the disease progression or treatment response can be monitored. However, interpreting these large MRI datasets requires significant expertise and demands extensive human resources. A tool for automatic diagnosis could address these challenges, offering high-quality, standardized and ubiquitously available screening. This grant proposal aims to leverage artificial intelligence (AI) to automate the detection, localization, and evaluation of GCA in extracranial vessels, enhancing diagnostic accuracy and efficiency. A data-driven framework will be developed which automatizes the pre-processing of MRI of the extracranial arteries, the detection of inflamed vessel segments and the disease quantification by segmentation of mural and perivascular contrast enhancement. For this purpose, a unique data archive will be assembled with cooperating GCA-centers in Germany and Switzerland. The data archive comprises contrast-enhanced T1-weighted 2D and 3D MRI data of patients with GCA and corresponding clinical biomarkers and histopathologic evaluation as ground truth. Within the pipeline, the MRI data will be automatically pre-processed to reduce the regions of interest to the scalp. A super- resolution model will be developed to ease the recognition of the coursed superficial temporal arteries within the abundant 2D datasets. Ultimately, the framework will detect vessel segments with signs of vasculitis and quantify disease severity by automatic segmentation and determination of the expansion of mural and perivascular enhancement. The overall aim of this grant proposal is the distribution of an AI-driven framework for automated detection of GCA in MRI, which was trained on multi-centered image data with intend for a high generalizability. With this, we aim to sustainably improve the diagnostics of giant cell arteritis in MRI and standardize it with high quality.

• DFG Programme: Research Grants
• International Connection: Switzerland
• Co-Investigators: Professor Dr. Fabian Bamberg; Professor Dr. Raoul Bergner
• Cooperation Partner: Professorin Dr. Sabine Adler