Interactive 3D segmentation and inspection of volumetric medial datasets

We propose an interactive method providing 3D real-time visualization of segmentation results while tuning some of the algorithmic parameters. Visual inspection in volume reduces the time spent in tuning cumbersome parameters and may increase accuracy in medical applications. To allow fast interaction, volume rendering is achieved by using 3D texture mapping. The output of the segmentation stage is then dynamically updated in the graphic pipeline through a color lookup table related to the tuned parameters. This technique enables our approach with immediate rendering of the user interaction during the segmentation. Isosurface methods and connectivity filters have been implemented with this technique. CT and MR modalities have been tested for anatomical structures extraction. For application in craniofacial surgical planning, measurements present improvement in accuracy and efficiency for 7 pathological cases. However, manual refinement is still necessary in order to realize clinical applicable 3D models.