Modeling surgical procedures for multimodal image-guided neurosurgery

Abstract

We present a model of surgical procedures that facilitates the management of multimodal information (i.e. anatomical and functional) in the context of multimodal image-guided craniotomies. We suggest that a surgical procedure be considered as a script consisting of a set of successive steps. Each step comprises a list of image entities extracted from multimodal images and relevant for the performance of this surgical step, and an action describing the purpose of the step. Each image entity has a role that may be different from step to step. This model is described by a UML class diagram and a textual description. We also present the results of a preliminary validation of the model performed on 15 clinical cases. The validation has shown the relevance of our model on the clinical cases studied and has confirmed the main assumptions we made to define the model. Even though further semantic validation is required, some benefits of this approach can be outlined: it improves management of multimodal information by describing when and why images are essential in the performance of the surgical act (e.g. place and role of images in the surgical script) and adapts visualization and interaction modes in neuronavigation. Our model should enhance preparation and guidance of surgery and lead to the development of simulation and even teaching tools.