We study new imaging techniques in CT and MRI for quantitative imaging of the spine. We develop automated evaluation algorithms including artificial intelligence, deep learning and biomechanic modeling to predict disease progression and get new pathophysiologic insights. Currently we focus on osteoporosis, back pain, degenerative spine diseases and inflammatory disorders like multiple sclerosis. We work with in-house datasets as well as with data from GNC and SHIP studies
We develop AI-based methods to automatically analyze and quantify MR images of the spine. We currently work on methods to translate MRI to CT to use existing labeling and segmentation methods. Our goal is to analyze large datasets from GNC to better understand sources of back pain.
To better understand back pain and surgical outcome, we combine quantitative parameters from different imaging modalities (X-rays, MRI, CT) in one biomechanical model, based on multibody simulations (MBS) on a large scale and finite element models (FEM/FCM) on a small scale.
We investigate different techniques to automatically assess spinal lesion load and its clinical relevance for the individual patient.
We closely collaborate with the group of Prof. Mühlau and Prof. Cohen-Adad to further increase functionaliy and robustness of the spinal cord toolbox.
Anduin is a freely
available web-based application able to fully automatically
segment the spine in clinical CT scans. Our spin-off "Bonescreen"
has further developed the segmentation code (contact
them directly for scientific collaborations) and is
dedicated to establish a fully automated opportunistic
osteoporosis diagnosis in clinical routine.
Try out Bonescreen!
Driving forward an interest for fully automated segmentation of the spine in the research community, we publicly released voxel-level-annotated CT data of more than 300 patients and organised the VerSe (Large Scale Vertebrae Segmentation) challenge series MICCAI 2019 and 2020. See the resulting MedIA paper (or ArXiv here)
Together with Bonescreen, we validate and develop automated techniques for opportunistic BMD measurements in any CT dataset. We also implemented automatic fracture detection and prediction algorithms using artificial intelligence. Currently, we focus on the spine.
Technische Universität München, Ismaninger Str. 22, 81675 München
Key highlights
Key highlights