This is an old revision of the document!
PyDBS is an automated image processing workflow for planning and postoperative assessment of deep brain stimulation interventions. The workflow is depicted in Fig. 1. It takes as input patient-specific data (i.e. patient images and patient clinical data) and generic models (i.e. an anatomical atlas, a model of the stereotactic frame and a model of the implanted electrodes) and provides as output a patient-specific model for planning and postoperative assessment of DBS surgery. This patient-specific model is composed of patient images, segmented anatomical structures (volumetric binary masks and triangular surface meshes) and geometrical transformations (registration matrices and deformation fields). All images, masks and meshes are mapped to a common reference space and fused in geometrical 3D scenes that can be readily visualized by the surgeon. The common reference space we adopt is the AC-PC space, defined by two anatomical brain landmarks: the anterior commissure (AC) and the posterior commissure (PC).
Three pipelines have been implemented: the inclusion pipeline, the preoperative pipeline and the postoperative pipeline. The inclusion pipeline, takes as input a T1-weighted and a T2-weighted MR image of the patient's head and an anatomical brain atlas. The T1-weighted image is used for an intensity-based segmentation of skin, brain and cortical sulci, and for an atlas-based segmentation of the brain ventricles and the basal ganglia. The T2-weighted image is registered to the reference AC-PC space for better visualization of deep brain structures. Since patient MR images are acquired several months before surgery, no tight time constraints are imposed to this pipeline. The preoperative pipeline takes as input a CT image of the patient's head fixed to the surgical stereotactic frame and a model of the stereotactic frame itself. The CT image is registered to the reference space and used both to segment the patient's skull and to detect the stereotactic frame on the image. Frame detection is used to compute a transformation that maps the frame space to the reference AC-PC space. Since the preoperative CT is acquired on the same day of the surgery, the processing time of this pipeline has direct impact on the overall duration of the surgical workflow and shall be thus minimized. The postoperative pipeline, takes as input a postoperative CT image of the patient's head and a geometrical model of the implanted electrode(s). The postoperative CT image is registered to the reference space and the electrode artifacts are segmented on the image. The geometrical electrode model is then fitted to the segmented artifacts to localize electrode contacts.