Abstract:

The focus of this work is the extraction and processing of information from bi-domain surgical video data to detect a vessel. The examined videos in the infrared and visible spectrum were recorded with a neurosurgical microscope during aneurysm procedures. The detection of a vessel is realized using the following three steps: Firstly, the automatic frame selection in the IR domain. Secondly, the segmentation of the vessel. Thirdly, the information matching and overlay using mutual information as metric. The methods were tested with four clinical video pairs and the highest average jaccard index achieved was 68%. The results show that the combination of information from multiple sensors is possible and enables a higher jaccard index than a single domain approach. It is now possible to use the geometric information about the vessel obtained from the indocyanine green (ICG) in combination with the RGB video over longer periods.