Challenge to be addressed
In 2D X-ray coronary angiography, multiple vessels can overlap due to projection of the 3D vascular structure onto a 2D plane, casuing ambiguous or falsely thickened vessel regions that degrade segmentation accuracy and morphological measurement reliability
2025-10-18
To improve my English writing skills, I`ve decided to write this post in English
There is a clearly darker area in the center of the segmentation result, caused by overlapping vessels in the projection.
Inaccurate 2D segmentation results leads to inaccurate 3D registation. Therefore, detecting overlapping vessels in the 2D segmentation result is essential.
I will address this problem using machine learning. I thick it can learn effectively even with a small amount of data.
The model must destinguish between self-overlap (the same vessel folding onto it self) and inter-vessel overlap (different vessels projected onto the same 2D position).