- 1997-2002. Neurosurgery department, Clermont Hospital
1999-2003 : La pitié Salpétrière Hospital, Paris
Expert knowledge information relate to indications of position, of form…, and can be clarified as fuzzy proposals (modeling the reflexions of a specialist) and/or as a brain atlas . Fusion thus takes place on heterogeneous and rather complementary data which are not necessarily expressed within the same formal framework (numerical for the images and the atlas, linguistics for expert knowledge). The interest of fusion is here to push back the decision (the segmentation of structures for example) until having sufficient information. This remark, which prevails in general for fusion, is remarkably illustrated in this type of problem, where the information of gray level or coarse position is not enough with the segmentation, but where the aggregation of the whole set of knowledge allows to infer a reliable decision.
The use of this type of method relates mainly to the segmentation of cerebral structures starting from images, atlas and symbolic data modeling expert knowledge. This segmentation can take place for the whole brain with, for example, a research of the hemispheres and cerebellum or stick more particularly to the detection of fine anatomical structures implied in specific pathologies. The clinical implications go then from the targeting of structures for a help to the surgical act to the follow-up of morphometric measurements of a pathological entity.
We applied this method :
- to the segmentation of the subthalamic nucleus within the framework of the treatment of the idiopathic Parkinson’s disease by electrostimulation
- to the segmentation of striatum