Developing a method for determining severity of hemodynamic coronary lesions from routine angiograms, thus sparing redundant interventions.
Vision Elements led the algorithmic and software development from PoC to working product. It created a 3D reconstruction of the vessel tree from a few X-ray views and extracted the vessels’ widths. Deep learning techniques were used for segmenting the vessels.
Challenges: Vessel movement (heart beats, breathing), noisy images, vessel cross-overs.
Keywords: medical, 3D, deep learning, detection