I-VESSEG:

Interactive and Collaborative Learning for Vessel Segmentation

I-VESSEG is an ANR JCJC funded project that aims to develop a novel interactive and collaborative learning-based framework to close the gap that hampers the use of 3D vessel segmentation tools in clinical routine.

The project will tackle three problems:

1. The difficulty to collect large representative training samples.
2. The burden associated to manual labelling.
3. The poor generalization and reproducibility of current vessel extraction approaches.

Clinical Use Cases

The final goal of I-VESSEG is to assure clinical translation. Its clinical value will be demonstrated in two clinical use cases:

• Multiple sclerosis (MS): I-VESSEG will be used to estimate the central vein sign, an imaging biomarker in MS which refers to a vein visualized inside a white matter lesion on susceptibility-weighted images.
• Stroke: I-VESSEG will assist the analysis of the cerebrovascular tree anatomy to detect intracranial stenosis, a major risk factor for stroke.