The grant received must be used to develop deep learning methods for the segmentation of large 3D images resulting from X-ray computed tomography.
This research is essential for analyzing images captured at synchrotron facilities like MAX IV in Lund, or laboratory facilities like DANFIX. Images taken at such facilities are captured in an attempt of revealing the internal structure of the sample; this may be a biopsy of nerve tissue, an eye of an insect, a bone sample, or a piece of fiber-reinforced polymer. The segmentation methods developed in STUDIOS will allow us to quantify the microstructure of the sample, whether it originates from a man-made material, a natural material, or a biological tissue.
“The research community working with X-ray imaging has not yet harvested the fruits of deep learning. This may seem surprising since deep learning solves extremely complex computer vision problems. But we still struggle to analyze large 3D images of materials and tissues. On top of that, each image captured at a synchrotron is a remarkable and very expensive accomplishment in itself. It is a waste, not to fully analyze it. With the grant by VILLUM FONDEN we will develop methods especially suitable for 3D X-ray images. This will allow us to analyze more such images, and to learn even more about the micro-structure of materials and tissues,” says Vedrana Andersen Dahl, associate professor, DTU Compute.
The project received DKK 1,999,696 from the VILLUM FONDEN.
Read more about the project: https://people.compute.dtu.dk/vand/studios/
