Many surgical procedures require highly precise localization, often of deeply buried structures, in order for the surgeon to extract targeted tissue with minimal damage to nearby structures. While methods such as MRI and CT are valuable for imaging and displaying the internal 3D structure of the body, the surgeon must still relate what he sees on the 3D display with the actual patient.
Traditional clinical practice often only utilizes 2D slices of MR or CT imagery, requiring the surgeon to mentally transform that information to the actual patient, thus there is a need for techniques to register 3D reconstructions of internal anatomy with the surgical field. Such image-guided surgical tools allow the surgeon to directly visualize important structures, and plan and act accordingly. Visualization methods include ``enhanced reality visualization'' [11], in which rendered internal structures are overlaid on the surgeon's field-of-view, and instrument tracking, in which medical instruments acting on the patient are localized and visualized in the 3D MR or CT imagery.
The key components of an accurate, reliable, image-guided surgery system are: creating accurate, detailed, patient-specific models of relevant anatomy for the surgical procedure; registering the models, and the corresponding imagery, to the patient; maintaining the registration throughout the surgical procedure; and tracking medical instruments in the surgical field in order to visualize them in the context of the MR/CT imagery and the reconstructed models.
We have developed a system which addresses these issues, primarily in neurosurgery. We had earlier reported our registration algorithms [11], and algorithmic testing of the system [10]. The important developments we report here are integration of tracking techniques, engineering of the system into an effective surgery tool, evaluation of the system's performance under control conditions, and initial experience of using the system on 70 cases in the operating room.