Cranio-maxillofacial surgery is a medical specialty focusing on facial and skull reconstruction. This surgery can help patients with such disorders as cleft palate, malformations of the upper or lower jaw, and problems with the facial skeleton due to injury. Intensive pre-operative planning is needed not only to ensure that the medical purposes of the surgery are achieved, but also to give patients a sense of what their faces will look like after the surgery is performed.
In their article “Mathematics in Facial Surgery,” Peter Deuflhard, Martin Weiser, and Stefan Zachow (of the Konrad Zuse Zentrum (ZIB), Berlin) describe the mathematical techniques they have used to assist cranio-maxillofacial surgeons to predict the outcomes of surgery. These techniques have proven to be quite successful in producing predictions that end up matching well the post-operative outcomes.
The first step in the planning paradigm for such surgery is to use medical imaging data of the patient to construct a 3-dimensional computer model, called the “virtual patient”. The second step, which is the one the article focuses on, uses the data to create a “virtual lab” in which various operative strategies can be tested. The last step is to play back to the patient the outcomes of the various strategies.
The second step in the paradigm requires modeling and solving partial differential equations (PDEs), which are equations that represent changing physical systems. One must identify which PDEs are appropriate for biomechanical modeling of soft facial tissue and bone. Standard methods for handling the equations need to be adapted for this particular application. One must also formulate ways to represent the interface between tissue and bone, as well as their interactions. Generally such PDEs cannot be solved exactly in closed form, so mathematics enters the picture once again to provide numerical techniques for producing approximate solutions.
With the “virtual patient” data as input, one can use the approximate solutions to generate an individualized model for that particular patient. The surgeons can then use the model as a “virtual lab” to predict the effects of surgical procedures and options, and patients can get a picture of approximately how they will look after the surgery.
The article by Deuflhard et al states that qualitative comparisons between the outcomes predicted by the model, and the actual surgical outcomes, have been surprisingly good. The authors have also made quantitative comparisons, by creating a post-operative model of the patient and comparing it quantitatively to the predicted outcome. They found a mean prediction error of between 1 and 1.5mm for the soft tissue, which they write “seems to be a fully acceptable result.”
“Even though biomechanical tissue modeling turns out to be a tough problem, we are already rather successful in predicting postoperative appearance from preoperative patient data,” the authors write. “For the surgeon, our computer assisted planning permits an improved preparation before the actual operation.”