Electromagnetic navigation in mandible reconstruction surgery

Abstract

In patients with oral cavity cancer invading the mandible, a segmental bone resection is performed and the original contour of the mandible is reconstructed with a free fibula flap. Virtual surgical planning is performed prior to surgery to determine the locations and orientations of the osteotomy planes on the mandible and fibula. Currently, patient-specific cutting guides are designed and three-dimensional printed to translate the virtual surgical plan to the patient in the operating room. However, these cutting guides are not ideal; they lack adaptability when the intraoperative situation is different than expected, e.g. due to tumor progression. Alternatively, surgical navigation could be used to translate the virtual surgical plan to the patient. To enable surgical navigation, accurate alignment of the preoperative imaging data, including the virtual surgical plan, and the patient is required, i.e. image-to-patient registration. In this thesis, a simple, accurate, and noninvasive registration method for electromagnetic navigation of the mandible is introduced and assessed.

Chapter 1 gives an introduction to the clinical background of mandible reconstruction surgery and the technical background of surgical navigation. The clinical problem, a potential solution and the thesis objectives are also discussed.

The systematic review in Chapter 2 gives an overview of currently used registration methods in navigated mandibular surgery: point registration, surface registration, hybrid registration, and computer vision based registration. The main conclusion of the review was that there is always a tradeoff between the usability, registration time, accuracy, and invasiveness of a registration method.

Chapter 3 introduces a simple and noninvasive registration method for mandible navigation: hybrid registration. This method consists of two steps: 1) point registration; performed for initialization using three anatomic landmarks on the mandible, and 2) surface registration; performed for optimization using the surgically exposed mandibular bone surface after removal of soft tissue.

In previous research in the NKI-AvL, an applicator was used to fixate an electromagnetic sensor to the mandible to track its movements during navigated surgery. The design of this applicator, however, enabled movement of the sensor in the applicator, which resulted in inaccurate navigation. Therefore, in Chapter 4, a renewed design for the sensor applicator is proposed.

In Chapter 5, the optimal approach for hybrid registration of the mandible is determined in phantom experiments. Different registration configurations, i.e. different surface point areas and number and configuration of surface points, were evaluated as well as registration with different patient anatomies. In all experiments, the target registration error (TRE) was below 2.0 mm, which meets the practical clinical requirements for mandible reconstruction surgery. The results suggest that only a small surface area of the mandible, marked by limited surface points, is required to obtain accurate registration.

Chapter 6 describes the preliminary results of hybrid registration of the mandible in four patients during surgery. Registration could be performed within on average 4.5 minutes. Mean TRE values of 3.4 mm for anatomic landmarks and 2.3 mm for cutting guide landmarks were obtained, indicating that the registration procedure should be further optimized to achieve clinically acceptable registration accuracy.

Chapter 7 provides an overall conclusion and future perspectives. Although the preliminary results of the patient study for mandible navigation are promising, the registration method should be further optimized and evaluated in more patients before implementation into clinical practice is possible. Ultimately, we want to use electromagnetic navigation to position a universal cutting guide during mandible reconstruction surgery. Multiple challenges still lie ahead before this can become reality in the NKI-AvL.