Medical Application

MedShapeNet – Medical 3D shapes for computer vision

Before the advent of deep learning, “shape” was typically used to describe objects. Today, modern (SOTA) algorithms in medical imaging largely differ from those in computer vision, which rely on voxel grids, networks, point clouds, and implicit surface models. This is evident in numerous shape-related publications at leading image processing conferences, as well as in the growing popularity of ShapeNet (approximately 51,300 models) and Princeton ModelNet (127,915 models). For the medical field, we present a large collection of anatomical shapes (e.g., bones, organs, vessels) and 3D models of surgical instruments, called MedShapeNet, which is designed to facilitate the transfer of data-driven image processing algorithms to medical applications and the adaptation of state-of-the-art (SOTA) image processing algorithms to medical problems. What makes this unique is that we model most of the shapes directly from image data of real patients. To date, MedShapeNet comprises 23 datasets containing more than 100,000 shapes paired with ground-truth annotations. Our data is freely accessible via a web interface and a Python application programming interface (API) and can be used for discriminative, reconstructive, and variational benchmarks, as well as for various applications in virtual, augmented, or mixed reality and in 3D printing. As examples, we present use cases in the areas of brain tumor classification, skull reconstruction, multi-class anatomical completion, education, and 3D printing. In the future, the data will be expanded and the interfaces improved.

The medical models are freely accessible at: https://medshapenet.ikim.nrw/

Publication

Li, J., Pepe, A., Gsaxner, C. et al. MedShapeNet — a Large-Scale dataset of 3D medical shapes for computer vision. arXiv (Cornell University) (2023). https://doi.org/10.48550/arXiv.2308.16139

Collaborating Partners

Gijs Luijten
Christina Gsaxner
Antonio Pepe
Behrus Puladi
Oliver Basu
Peter Hoyer
Jens Kleesiek
Jan Egger

Contact

Jianning Li (Jianning.Li@uk-essen.de)

Subscribe to the newsletter

Newsletter abonnieren