)
A research team around 2020’s Award Winner PD Dr. Philipp Vollmuth and international cooperation partners revealed how artificial intelligence (AI) can be used in medical imaging to reduce the use of gadolinium as a contrast agent in magnetic resonance imaging (MRI). An MRI exam is widely used in hospitals and clinics for medical diagnosis, staging and follow-up of various diseases and a significant number of these exams are performed with Gadolinium based contrast agents (GBCAs) to improve the visualization of pathological processes. However, one of the most controversial issues in radiology in recent years has been the use of GBCA and the European Medical Agency has recommended restrictions and suspensions for some GBCAs, based on recent studies demonstrating gadolinium deposition in the brain after repeated GBCA administration with yet unknown clinical significance. Consequently, there is a great scientific interest in exploring potential alternatives to lower the need of administering GBCAs during MRI.
In our proof-of-principle study we developed and validated a novel AI based model that allows virtual synthesis of information (tissue contrasts) that would otherwise only have been visible with administration of GBCAs. By leveraging large-scale MRI data from over 200 institutions worldwide with more than 2000 patients we have demonstrated the feasibility and clinical validity of our approach within the context of brain tumors (see Figure 1). Lowering the need of administering GBCAs has the potential to benefit a great number of patients, including pediatric and neonatal patients, as well as patients with neurological diseases such as brain tumors who frequently undergo repetitive contrast-enhanced MRI exams. Our current research efforts are therefore directed towards further clinical validation and extension of these promising developments to other disease entities beyond brain tumors.
Figure 1: True post-contrast T1-weighted MRI sequences with administration of GBCAs (top) vs. synthetic post-contrast T1-weighted MRI sequences without administration of GBCAs (bottom) from two timepoints (baseline and follow-up MRI) in three illustrative patients with brain tumors. Each cases demonstrates high spatial and temporal agreement in tumor burden between true and synthetic post-contrast T1-weighted MRI.
The author Philipp Vollmuth is grateful to the Aventis Foundation for the Life Sciences Bridge Award which has importantly contributed to the success of this project. In particular, the award enabled to acquire one of the world’s most advanced AI systems thereby providing state-of-the-art computational performance for accelerating the scientific progress within our research.
The full article can be found at: Jayachandran Preetha C, Meredig H, Brugnara G, Mahmutoglu MA, Foltyn M, IsenseeF, Kessler T, Pflüger I, Schell M, Neuberger U, Petersen J, Wick A, Heiland S, Debus J, Platten M, Idbaih A, BrandesAA, Winkler F, van den Bent MJ, Nabors B, Stupp R, Maier-Hein KH, Gorlia T, Tonn JC, Weller M, Wick W, BendszusM, Vollmuth P. Deep-learning-based synthesis of post-contrast T1-weighted MRI for tumour response assessment inneuro-oncology: a multicentre, retrospective cohort study. Lancet Digit Health. 2021 Dec;3(12):e784-e794