A New Software Environment for 3D-Time Series Analysis

Introduction

In this chapter, the authors present and validate a toolchain that allows the analysis of changes in a patient or specimen based on a series of DICOM CT images taken at different times by using image segmentation, voxel based rigid and non-rigid registration, as well as 3D visualization. This time series analysis of medical images enables the full three-dimensional comparison of two or more data sets of the same person or specimen, e.g., the comparison of pre- and post treatment CT images. Such analysis helps to understand the underlying processes, that arise, for instance, from treatment.

One prominent example of treatment inducing changes of shape in patients is distraction osteogenesis (a method to correct severe maxillary hypoplasia or retrusion). In clinical practice, this planning is often based on computer tomography (CT) scans and the surgeon’s experience. Planning these procedures can be difficult if complex three-dimensional changes are to be performed (Berti et al., 2004), and as a first step towards an improved treatment planning, a thorough understanding of what is achieved and how the different structures have been moved is needed. Here a better understanding of the structural changes induced by the surgical therapy is desirable as a first step. Given a thorough understand of the procedure, a superior treatment planning and outcome could be achieved.

In dentistry, the scope of the presented toolchain generally ranges from pre- and postoperative oral surgery images (orthognathic surgery, trauma surgery) to endodontic and orthodontic treatment as well as the analysis of growth. The new tool is already part of our operation planning and visualization software. We concentrated on the time series analysis of routinely acquired pre- and postoperative CT-images of patients who were treated via maxillary distraction osteogenesis as no studies regarding the analysis of the complex three-dimensional mid-facial movements are available (Hierl et al., 2005). Furthermore, incorrect treatment planning may lead to a malpositioned midface and may necessitate further surgical intervention. At the end of this chapter we give some examples of the results of the time series analysis performed. Finally validation of the registration methods is performed. This means showing that our registration algorithm applied consistently succeeds with an average error acceptable for the application. In the near future the toolchain will be improved and integrated into a multifunctional planning tool. Furthermore, the range of data files will be enlarged. By now only DICOM files from CT or CBT (cone beam tomography) can be loaded. In the future polygon-based surfaces should be included, too.